[{"data":1,"prerenderedAt":4111},["ShallowReactive",2],{"i-kinnu:logo":3,"i-kinnu:origami-folding":8,"pathway-technology-artificial-intelligence":12,"i-lucide:chevron-right":4106,"i-lucide:tag":4109},{"left":4,"top":4,"width":5,"height":5,"rotate":4,"vFlip":6,"hFlip":6,"body":7},0,27,false,"\u003Cg fill=\"none\">\u003Cpath d=\"M0.046875 1.05555C0.046875 1.03541 0.048197 1.01579 0.0507438 0.996728C0.0987149 0.438619 0.586845 0 1.18194 0H25.4398C26.451 0 26.9575 1.171 26.2424 1.85585L15.7301 11.9243L1.31574 0.903476C1.17475 0.79568 1.01137 0.761884 0.859586 0.784111L26.2936 25.1441C27.0086 25.829 26.5022 27 25.4909 27H1.18194C0.555061 27 0.046875 26.5133 0.046875 25.9129V1.05555Z\" fill=\"currentColor\"/>\u003C/g>",{"left":4,"top":4,"width":9,"height":10,"rotate":4,"vFlip":6,"hFlip":6,"body":11},1000,236,"\u003Cg fill=\"none\">\u003Cpath fill-rule=\"evenodd\" clip-rule=\"evenodd\"\n    d=\"M193.68 38.2238C195.994 38.2238 197.87 40.0989 197.87 42.412V231.812C197.87 234.125 195.994 236 193.68 236H4.19013C1.87603 236 2.02305e-07 234.125 0 231.812V42.412C-2.02305e-07 40.0989 1.87603 38.2238 4.19013 38.2238H193.68ZM111.76 89.0072C111.685 87.9474 110.572 87.2905 109.608 87.7376L96.8872 93.641C95.7786 94.1554 95.702 95.7016 96.7545 96.3225L101.579 99.167C94.7045 109.365 90.5733 122.892 90.5732 137.642C90.5733 154.323 95.8569 169.439 104.416 179.945C105.301 181.032 106.9 181.196 107.987 180.311C109.075 179.426 109.238 177.828 108.353 176.741C100.621 167.25 95.6522 153.305 95.6521 137.642C95.6522 123.661 99.6138 111.051 105.963 101.754L110.456 104.403C111.508 105.024 112.826 104.21 112.74 102.991L111.76 89.0072ZM9.63194 136.286C9.14864 136.286 8.75684 136.678 8.75684 137.161C8.7569 137.644 9.14868 138.035 9.63194 138.035H17.2161C17.6993 138.035 18.0912 137.644 18.0912 137.161C18.0912 136.678 17.6994 136.286 17.2161 136.286H9.63194ZM22.6813 136.286C22.198 136.286 21.8062 136.678 21.8062 137.161C21.8063 137.644 22.1981 138.035 22.6813 138.035H30.2655C30.7487 138.035 31.1406 137.644 31.1406 137.161C31.1406 136.678 30.7488 136.286 30.2655 136.286H22.6813ZM35.7464 136.286C35.2631 136.286 34.8713 136.678 34.8713 137.161C34.8713 137.644 35.2631 138.035 35.7464 138.035H44.4973C44.9805 138.035 45.3724 137.644 45.3724 137.161C45.3724 136.678 44.9806 136.286 44.4973 136.286H35.7464ZM49.9977 136.286C49.5144 136.286 49.1226 136.678 49.1226 137.161C49.1226 137.644 49.5144 138.035 49.9977 138.035H57.5819C58.0651 138.035 58.4569 137.644 58.457 137.161C58.457 136.678 58.0651 136.286 57.5819 136.286H49.9977ZM63.0783 136.286C62.595 136.286 62.2032 136.678 62.2032 137.161C62.2033 137.644 62.5951 138.035 63.0783 138.035H70.6625C71.1457 138.035 71.5375 137.644 71.5376 137.161C71.5376 136.678 71.1457 136.286 70.6625 136.286H63.0783ZM76.1277 136.286C75.6444 136.286 75.2526 136.678 75.2526 137.161C75.2527 137.644 75.6445 138.035 76.1277 138.035H83.7119C84.1951 138.035 84.5869 137.644 84.587 137.161C84.587 136.678 84.1951 136.286 83.7119 136.286H76.1277ZM102.266 136.286C101.782 136.286 101.39 136.678 101.39 137.161C101.391 137.644 101.782 138.035 102.266 138.035H109.85C110.333 138.035 110.725 137.644 110.725 137.161C110.725 136.678 110.333 136.286 109.85 136.286H102.266ZM115.338 136.286C114.855 136.286 114.463 136.678 114.463 137.161C114.463 137.644 114.855 138.035 115.338 138.035H122.923C123.406 138.035 123.798 137.644 123.798 137.161C123.798 136.678 123.406 136.286 122.923 136.286H115.338ZM128.403 136.286C127.92 136.286 127.528 136.678 127.528 137.161C127.528 137.644 127.92 138.035 128.403 138.035H135.988C136.471 138.035 136.863 137.644 136.863 137.161C136.863 136.678 136.471 136.286 135.988 136.286H128.403ZM141.468 136.286C140.985 136.286 140.593 136.678 140.593 137.161C140.593 137.644 140.985 138.035 141.468 138.035H149.053C149.536 138.035 149.928 137.644 149.928 137.161C149.928 136.678 149.536 136.286 149.053 136.286H141.468ZM154.541 136.286C154.058 136.286 153.666 136.678 153.666 137.161C153.666 137.644 154.058 138.035 154.541 138.035H162.125C162.609 138.035 163 137.644 163.001 137.161C163.001 136.678 162.609 136.286 162.125 136.286H154.541ZM167.614 136.286C167.131 136.286 166.739 136.678 166.739 137.161C166.739 137.644 167.131 138.035 167.614 138.035H175.198C175.681 138.035 176.073 137.644 176.073 137.161C176.073 136.678 175.681 136.286 175.198 136.286H167.614ZM180.671 136.286C180.188 136.286 179.796 136.678 179.796 137.161C179.796 137.644 180.188 138.035 180.671 138.035H188.255C188.739 138.035 189.13 137.644 189.131 137.161C189.131 136.678 188.739 136.286 188.255 136.286H180.671Z\"\n    fill=\"currentColor\" />\n  \u003Cpath fill-rule=\"evenodd\" clip-rule=\"evenodd\"\n    d=\"M444.85 38.2277C447.164 38.2277 449.04 40.1028 449.04 42.4159V132.928C449.04 135.241 447.164 137.116 444.85 137.116H255.36C253.046 137.116 251.17 135.241 251.17 132.928V42.4159C251.17 40.1028 253.046 38.2277 255.36 38.2277H444.85ZM361.96 125.388C361.618 125.046 361.064 125.046 360.722 125.388L354.534 131.572C354.192 131.914 354.192 132.468 354.534 132.81C354.876 133.151 355.43 133.151 355.772 132.81L361.96 126.624C362.301 126.283 362.301 125.73 361.96 125.388ZM371.047 116.311C370.705 115.969 370.15 115.969 369.809 116.311L364.446 121.671C364.104 122.012 364.104 122.567 364.446 122.908C364.788 123.249 365.342 123.25 365.684 122.908L371.047 117.548C371.388 117.207 371.388 116.652 371.047 116.311ZM380.124 107.246C379.782 106.904 379.227 106.904 378.885 107.246L373.523 112.606C373.181 112.948 373.181 113.502 373.523 113.844C373.864 114.185 374.419 114.185 374.761 113.844L380.124 108.483C380.465 108.142 380.465 107.587 380.124 107.246ZM385.736 65.8841C385.891 64.6727 384.622 63.7845 383.536 64.3434L371.069 70.7636C370.124 71.2504 369.96 72.5334 370.752 73.2424L381.2 82.5938C382.11 83.4081 383.561 82.8672 383.717 81.6557L384.393 76.3725C391.143 77.1933 398.567 80.7709 404.771 86.9711C411.124 93.3213 414.726 100.952 415.43 107.827C415.573 109.221 416.819 110.236 418.214 110.093C419.609 109.95 420.624 108.703 420.481 107.309C419.644 99.1317 415.435 90.4514 408.362 83.3817C401.466 76.489 393.038 72.3185 385.038 71.338L385.736 65.8841ZM389.2 98.1733C388.859 97.8319 388.304 97.8318 387.962 98.1733L382.6 103.534C382.258 103.875 382.258 104.429 382.6 104.771C382.941 105.112 383.496 105.112 383.838 104.771L389.2 99.4108C389.542 99.0693 389.542 98.5149 389.2 98.1733ZM398.262 89.1047C397.92 88.7633 397.365 88.7632 397.024 89.1047L391.661 94.4649C391.319 94.8065 391.319 95.3608 391.661 95.7024C392.002 96.0436 392.557 96.0438 392.899 95.7024L398.262 90.3421C398.603 90.0007 398.603 89.4463 398.262 89.1047ZM416.431 70.9616C416.089 70.6202 415.534 70.6201 415.193 70.9616L409.83 76.3218C409.488 76.6634 409.488 77.2177 409.83 77.5592C410.172 77.9005 410.726 77.9007 411.068 77.5592L416.431 72.199C416.772 71.8575 416.772 71.3032 416.431 70.9616ZM425.508 61.891C425.166 61.5496 424.611 61.5495 424.27 61.891L418.907 67.2512C418.565 67.5928 418.565 68.1471 418.907 68.4887C419.249 68.8299 419.803 68.8301 420.145 68.4887L425.508 63.1284C425.849 62.787 425.849 62.2326 425.508 61.891ZM434.569 52.8146C434.227 52.4731 433.673 52.4731 433.331 52.8146L427.968 58.1748C427.626 58.5163 427.627 59.0706 427.968 59.4122C428.31 59.7534 428.864 59.7537 429.206 59.4122L434.569 54.052C434.91 53.7105 434.91 53.1562 434.569 52.8146ZM443.638 43.7479C443.296 43.4065 442.742 43.4064 442.4 43.7479L437.037 49.1081C436.695 49.4496 436.696 50.004 437.037 50.3455C437.379 50.6868 437.933 50.687 438.275 50.3455L443.638 44.9853C443.98 44.6438 443.979 44.0895 443.638 43.7479Z\"\n    fill=\"currentColor\" />\n  \u003Cpath fill-rule=\"evenodd\" clip-rule=\"evenodd\"\n    d=\"M684.066 38.2277C687.798 38.2281 689.667 42.7391 687.027 45.3773L596.473 135.889C595.687 136.675 594.621 137.116 593.51 137.116H506.335C504.021 137.116 502.145 135.241 502.145 132.928V42.4159C502.145 40.1028 504.021 38.2277 506.335 38.2277H684.066ZM514.603 124.566C514.261 124.224 513.707 124.224 513.365 124.566L507.178 130.751C506.836 131.093 506.836 131.646 507.178 131.988C507.519 132.329 508.073 132.329 508.415 131.988L514.603 125.803C514.945 125.462 514.945 124.908 514.603 124.566ZM523.689 115.491C523.348 115.15 522.794 115.15 522.452 115.491L517.09 120.852C516.748 121.193 516.748 121.747 517.09 122.088C517.431 122.43 517.985 122.43 518.327 122.088L523.689 116.728C524.031 116.386 524.031 115.833 523.689 115.491ZM532.102 65.8295C530.707 65.6872 529.46 66.7017 529.318 68.0957C529.175 69.4896 530.189 70.7355 531.584 70.8787C538.463 71.5825 546.096 75.1826 552.45 81.5329C558.723 87.8037 562.312 95.3226 563.079 102.13L557.738 102.392C556.518 102.452 555.865 103.855 556.607 104.827L565.115 115.969C565.76 116.814 567.051 116.751 567.611 115.847L574.992 103.928C575.635 102.889 574.848 101.555 573.628 101.615L568.161 101.882C568.161 101.878 568.162 101.874 568.161 101.871C567.324 93.6931 563.114 85.0124 556.041 77.9425C548.968 70.873 540.283 66.6668 532.102 65.8295ZM532.766 106.421C532.425 106.079 531.871 106.079 531.529 106.421L526.166 111.781C525.825 112.123 525.825 112.676 526.166 113.018C526.508 113.359 527.062 113.359 527.403 113.018L532.766 107.657C533.108 107.316 533.108 106.762 532.766 106.421ZM541.843 97.3445C541.501 97.003 540.948 97.003 540.606 97.3445L535.243 102.705C534.901 103.046 534.902 103.6 535.243 103.941C535.585 104.283 536.139 104.283 536.48 103.941L541.843 98.5809C542.185 98.2393 542.185 97.686 541.843 97.3445ZM550.92 88.2778C550.578 87.9363 550.025 87.9363 549.683 88.2778L544.32 93.638C543.978 93.9796 543.978 94.5329 544.32 94.8745C544.662 95.2161 545.215 95.2161 545.557 94.8745L550.92 89.5142C551.262 89.1727 551.262 88.6193 550.92 88.2778ZM569.066 70.1405C568.724 69.799 568.17 69.7991 567.829 70.1405L562.466 75.5008C562.124 75.8423 562.124 76.3956 562.466 76.7372C562.808 77.0788 563.361 77.0788 563.703 76.7372L569.066 71.377C569.407 71.0354 569.407 70.4821 569.066 70.1405ZM578.143 61.0699C577.801 60.7284 577.247 60.7285 576.906 61.0699L571.543 66.4302C571.201 66.7717 571.201 67.3251 571.543 67.6666C571.885 68.0082 572.438 68.0082 572.78 67.6666L578.143 62.3064C578.484 61.9648 578.484 61.4115 578.143 61.0699ZM587.219 51.9896C586.878 51.6481 586.324 51.6481 585.982 51.9896L580.62 57.3498C580.278 57.6914 580.278 58.2447 580.62 58.5863C580.961 58.9279 581.515 58.9279 581.857 58.5863L587.219 53.2261C587.561 52.8845 587.561 52.3312 587.219 51.9896ZM596.288 42.9249C595.947 42.5833 595.392 42.5833 595.05 42.9249L589.689 48.2851C589.347 48.6267 589.347 49.18 589.689 49.5216C590.03 49.863 590.584 49.8631 590.926 49.5216L596.288 44.1613C596.63 43.8198 596.63 43.2664 596.288 42.9249Z\"\n    fill=\"currentColor\" />\n  \u003Cpath fill-rule=\"evenodd\" clip-rule=\"evenodd\"\n    d=\"M850.814 38.2277C854.547 38.2281 856.416 42.739 853.777 45.3773L763.223 135.889C762.437 136.674 761.371 137.116 760.26 137.116H673.176C669.443 137.116 667.574 132.605 670.213 129.966L760.768 39.4544C761.554 38.6692 762.62 38.2277 763.731 38.2277H850.814ZM761.338 121.8C760.855 121.8 760.463 122.191 760.463 122.674V131.13H762.213V122.674C762.213 122.191 761.821 121.8 761.338 121.8ZM761.338 108.971C760.855 108.971 760.463 109.363 760.463 109.846V118.301H762.213V109.846C762.213 109.363 761.821 108.971 761.338 108.971ZM761.338 96.1402C760.855 96.1406 760.463 96.5321 760.463 97.0149V105.47H762.213V97.0149C762.213 96.532 761.821 96.1404 761.338 96.1402ZM782.263 71.887C781.043 71.951 780.395 73.3571 781.139 74.3257L784.474 78.6631C779.115 82.951 771.242 85.7443 762.35 85.7444C753.366 85.7442 745.421 82.8944 740.059 78.5305C738.972 77.6461 737.373 77.8099 736.488 78.8961C735.602 79.983 735.766 81.582 736.853 82.467C743.231 87.6574 752.348 90.8207 762.35 90.8209C772.209 90.8208 781.205 87.746 787.568 82.6884L790.833 86.9341C791.577 87.9025 793.103 87.6391 793.479 86.4767L797.791 73.138C798.118 72.127 797.33 71.1017 796.268 71.1566L782.263 71.887ZM761.338 70.4847C760.855 70.4851 760.463 70.8767 760.463 71.3594V79.8147H762.213V71.3594C762.213 70.8766 761.821 70.485 761.338 70.4847ZM761.338 57.656C760.855 57.6564 760.463 58.048 760.463 58.5307V66.986H762.213V58.5307C762.213 58.0479 761.821 57.6563 761.338 57.656ZM761.338 44.8293C760.855 44.8297 760.463 45.2212 760.463 45.704V54.1592H762.213V45.704C762.213 45.2211 761.821 44.8295 761.338 44.8293Z\"\n    fill=\"currentColor\" />\n  \u003Cpath\n    d=\"M995.759 38.2277C999.53 38.228 1001.42 42.5171 998.752 45.0253L959.55 81.9005L905.796 41.5363C905.271 41.1418 904.662 41.0182 904.096 41.0994L997.485 130.319C1000.15 132.828 998.262 137.116 994.491 137.116H905.298C902.96 137.116 901.065 135.333 901.065 133.134V42.0941C901.065 42.0204 901.07 41.9483 901.079 41.8786C901.258 39.8345 903.079 38.2277 905.298 38.2277H995.759Z\"\n    fill=\"currentColor\" />\n  \u003Cpath\n    d=\"M505.873 0C506.657 4.57042e-05 507.307 0.195499 507.823 0.587023C508.338 0.969046 508.596 1.53802 508.596 2.29251C508.596 2.76034 508.467 3.19015 508.209 3.58162C507.951 3.96344 507.497 4.26401 506.848 4.48361V4.54114C507.65 4.67487 508.205 4.96191 508.51 5.4012C508.816 5.83087 508.969 6.31772 508.969 6.86193C508.969 7.74056 508.672 8.41851 508.08 8.89604C507.497 9.38304 506.733 9.62731 505.787 9.62738C504.861 9.62738 504.158 9.42172 503.68 9.0111C503.212 8.60054 502.935 8.08005 502.849 7.44993L503.881 7.10571L503.924 7.24028C504.035 7.54934 504.211 7.82925 504.454 8.07986C504.731 8.36635 505.166 8.50986 505.758 8.50989C506.465 8.50989 506.943 8.32772 507.191 7.9648C507.449 7.6019 507.579 7.20078 507.579 6.7615C507.579 6.2173 507.378 5.80683 506.977 5.52992C506.585 5.25295 505.93 5.10026 505.013 5.07161V4.15402C505.901 4.12537 506.489 3.92484 506.776 3.55237C507.062 3.18009 507.206 2.82242 507.206 2.47876C507.206 1.62801 506.752 1.17539 505.845 1.12237L505.658 1.11749C505.467 1.11752 505.242 1.14605 504.985 1.2033C504.736 1.25105 504.511 1.3274 504.31 1.43245L504.081 2.56457L503.05 2.44951L503.322 0.687461C503.666 0.49653 504.068 0.33454 504.526 0.200875C504.985 0.0671945 505.434 0 505.873 0Z\"\n    fill=\"currentColor\" />\n  \u003Cpath\n    d=\"M905.727 2.30616L904.638 2.4066L904.466 1.26083H901.428V3.72497C901.533 3.71544 901.643 3.71034 901.757 3.71034H902.086C902.755 3.71034 903.386 3.78668 903.979 3.93949C904.58 4.09229 905.068 4.38363 905.44 4.8132C905.822 5.23335 906.014 5.84949 906.014 6.66106C906.014 7.64468 905.722 8.38068 905.14 8.86776C904.557 9.36434 903.783 9.6127 902.818 9.61275C901.91 9.61275 901.213 9.40711 900.725 8.99648C900.248 8.59544 899.96 8.08007 899.865 7.44993L900.911 7.10571C901.007 7.49723 901.203 7.8271 901.499 8.09449C901.795 8.37131 902.211 8.50985 902.746 8.50989C903.395 8.50989 903.869 8.33787 904.165 7.99405C904.461 7.65981 904.609 7.22507 904.609 6.69031C904.609 5.87861 904.337 5.3625 903.792 5.14279C903.248 4.91361 902.612 4.79958 901.886 4.79955C901.695 4.79955 901.489 4.80365 901.27 4.8132C901.059 4.82275 900.854 4.83701 900.653 4.85611L900.224 4.44071V0.143343H905.569L905.727 2.30616Z\"\n    fill=\"currentColor\" />\n  \u003Cpath fill-rule=\"evenodd\" clip-rule=\"evenodd\"\n    d=\"M765.49 6.04576H766.966L766.837 7.14862H765.49V9.48404H764.185V7.14862H759.857L759.713 6.04576L762.909 0.143343H765.49V6.04576ZM760.96 6.04576H764.185V1.26083H763.541L760.96 6.04576Z\"\n    fill=\"currentColor\" />\n  \u003Cpath d=\"M4.80573 6.47481H6.41154V7.60693H1.81068V6.47481H3.50235V1.27546H1.81068V0.143343H4.80573V6.47481Z\"\n    fill=\"currentColor\" />\n  \u003Cpath\n    d=\"M254.359 0C255.353 0 256.055 0.239186 256.466 0.716715C256.877 1.18447 257.083 1.68072 257.083 2.20573C257.083 2.85516 256.849 3.44346 256.38 3.96875C255.912 4.49397 255.348 4.96638 254.689 5.38657C254.039 5.79717 253.437 6.15968 252.883 6.47481H256.423L256.538 5.42948L257.599 5.51529L257.426 7.60693H251.407L251.292 6.58987C252.582 5.73032 253.638 4.98523 254.46 4.35489C255.281 3.71509 255.693 3.05632 255.693 2.37832C255.693 1.53787 255.166 1.11749 254.115 1.12237L254.115 1.11749C253.924 1.11754 253.695 1.14604 253.427 1.2033C253.16 1.25104 252.916 1.32238 252.697 1.41783L252.467 2.47876L251.45 2.3637L251.707 0.60165C252.118 0.401088 252.563 0.253475 253.041 0.15797C253.519 0.0529708 253.958 1.99446e-05 254.359 0Z\"\n    fill=\"currentColor\" />\u003C/g>",{"id":13,"data":14,"type":15,"maxContentLevel":28,"version":29,"tiles":30},"eed5242a-ac86-4bd7-b401-a18f4116b837",{"type":15,"title":16,"tagline":17,"description":17,"featureImageSquare":18,"baseColor":19,"emoji":20,"shapePreference":4,"allowContentSuspension":21,"allowContentEdits":21,"editorsChoice":21,"accreditations":22,"certificatePriceLevel":27,"certificationTitle":16},8,"Artificial Intelligence","Welcome to the bleeding edge of computer-led cognition","0346ec71-b214-4959-b6e0-35229911ae94","#79909C","🤖",true,[23],{"authority":24,"wasCpdTill":25,"previousCpdCreditMinutes":26},1,"2025-12-31T00:00:00Z",120,3,9,10,[31,193,641,1030,1632,2074,2507,2886,3196,3474,3931],{"id":32,"data":33,"type":28,"maxContentLevel":36,"version":37,"orbs":38},"fb165e19-1b07-4bde-ac46-93eccbfd9a73",{"type":28,"title":34,"tagline":35},"What is AI?","Let's get started...",7,4,[39,105],{"id":40,"data":41,"type":42,"version":27,"maxContentLevel":36,"summaryPage":44,"introPage":51,"pages":57},"fe860c8e-c8c1-42ab-b15f-a50fdfb93bf0",{"type":42,"title":43},2,"Introducing AI",{"id":45,"data":46,"type":27,"maxContentLevel":27,"version":24},"24e2c322-bbf2-475e-ab33-f0f8be924d43",{"type":27,"summary":47},[48,49,50],"The Ancient Greeks spoke of Talos, a giant bronze automaton","Talos was only a myth, but in the modern day AI has become a reality","AI refers to machines which are capable of human-like cognitive processes",{"id":52,"data":53,"type":29,"maxContentLevel":27,"version":24},"3bff1402-d221-4eb8-9bea-0da660a3928c",{"type":29,"intro":54},[55,56],"What is the official definition of Artificial Intelligence?","Why is AI relevant to almost every person on the planet?",[58,76,81],{"id":59,"data":60,"type":24,"maxContentLevel":27,"version":42,"reviews":63},"cbfe4d38-6fef-4ae7-89b9-34b049e7f524",{"type":24,"contentRole":42,"markdownContent":61,"audioMediaId":62},"The Ancient Greeks spoke of Talos: an automated giant, more than thirty meters tall, with a body made entirely of bronze. He clanged and crashed with every massive step, while his forehead glittered in the sun.\n\nTalos had a single purpose. To protect the island of Crete. Every single day, he circled it three times, leaving footprints on the white-sanded shores. If he spotted an enemy, he would reach for a boulder. With a groan of effort, he raised it overhead, then launched it at the incoming threat.\n\nTalos, of course, was just a myth. A robot like this was well beyond what the Ancient Greeks could create. But in the modern world, things are changing. The AI revolution has arrived.\n\n![Graph](image://d4e39639-c168-4436-b10b-fce8a2c473eb \"An AI generated image.\")","5d2cce85-3d1b-446f-9c63-1279b5bbe18b",[64],{"id":65,"data":66,"type":67,"version":24,"maxContentLevel":27},"a126a0f0-bfed-4e83-904b-bb3701c371dd",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":68,"multiChoiceCorrect":70,"multiChoiceIncorrect":72,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},11,[69],"According to the Ancient Greeks, what was the primary purpose of Talos?",[71],"To protect the island of Crete",[73,74,75],"To conquer other islands","To serve the gods","To destroy the island of Crete",{"id":77,"data":78,"type":24,"maxContentLevel":27,"version":24},"cb178d1b-0821-4646-8068-6b5de987872f",{"type":24,"contentRole":42,"markdownContent":79,"audioMediaId":80},"AI stands for Artificial Intelligence. In general, this term is used to describe a machine or system which is capable of performing **human-like cognitive processes**. Examples of this include problem solving, logical reasoning, creative thinking, and learning.\n\nTalos (if he'd existed) would have fit this definition of AI. He used **logical reasoning** to determine whether an approaching ship was an enemy. He **learned** from his encounters, and if something went wrong, he used **creative thinking** to overcome it.\n\nAn AI can often perform these processes significantly faster than a human. And that's a pretty big deal. From sorting data, to guiding vehicles, to writing documents, to generating art, this is the first time in history that human-like processes can be performed at computer-like speeds.","a4a3e567-c32f-45f1-9f46-78c8a88adca7",{"id":82,"data":83,"type":24,"maxContentLevel":36,"version":42,"reviews":86},"ab78e65e-a191-423d-8f13-60c974fc3751",{"type":24,"contentRole":42,"markdownContent":84,"audioMediaId":85},"The rise of AI, in the modern world, is relevant to almost every person on the planet. No matter who you are, no matter where you live, AI will impact your life.\n\nThis might sound a bit daunting. And there are certainly some dangers to be aware of. But the AI revolution is also opening up a universe of new opportunities.\n\nThat's why it's so important for us to learn about these new technologies. And who better to guide us than Mehdi Ghissassi, the Director of Product at Google DeepMind, one of the world's leading Artificial Intelligence projects.\n\nThis pathway was written in collaboration between Kinnu and Mehdi – plus some help from AI along the way. By the time you're done, you'll be ready to thrive in this new, AI-driven world.\n\n![Graph](image://acd5847c-a7a1-4e75-a200-c02737ec191f \"Welcome to the future. Image: Toyota Kaikan (CC BY-SA 3.0) \u003Chttps://creativecommons.org/licenses/by-sa/3.0>, via Wikimedia Commons\")","ac1d32de-6c93-406f-9f72-cfc3294509e7",[87,94],{"id":88,"data":89,"type":67,"version":42,"maxContentLevel":36},"ad03a94f-0feb-4269-b24e-0ae2babd8870",{"type":67,"reviewType":24,"evolvingBehavior":24,"spacingBehaviour":24,"activeRecallQuestion":90,"activeRecallAnswers":92},[91],"'Artificial Intelligence' describes a machine or system which is capable of what?",[93],"Human-like cognitive processes",{"id":95,"data":96,"type":67,"version":24,"maxContentLevel":27},"c2b518fa-8868-4a0c-9a1c-09f5598456a9",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":97,"multiChoiceCorrect":99,"multiChoiceIncorrect":101,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[98],"What is a key advantage of AI over humans?",[100],"It can perform processes faster than humans",[102,103,104],"It can think more creatively than humans","It can understand emotions better than humans","It can make decisions more ethically than humans",{"id":106,"data":107,"type":42,"version":42,"maxContentLevel":27,"summaryPage":109,"introPage":117,"pages":123},"f427317b-598f-474a-a90f-02be9b37ef66",{"type":42,"title":108},"Different types of AI",{"id":110,"data":111,"type":27,"maxContentLevel":27,"version":42},"c964b8e5-8aaf-42d4-9eb5-814e9c4e1732",{"type":27,"summary":112},[113,114,115,116],"Artificial Narrow Intelligence (ANI) has cognitive abilities which are less than those of a human","Artificial General Intelligence (AGI) has cognitive abilities which are the same as those of a human","Artificial Super Intelligence (ASI) has cognitive abilities which are greater than those of a human","ANI already exists, but AGI and ASI are currently hypothetical",{"id":118,"data":119,"type":29,"maxContentLevel":27,"version":24},"d5c9585e-fac4-4611-b2f4-8b578dfe19ed",{"type":29,"intro":120},[121,122],"What are the three different types of AI?","Are modern AI models smarter than human beings?",[124,129,144,159],{"id":125,"data":126,"type":24,"maxContentLevel":27,"version":24},"b9a0b1c9-4af7-45e4-84b0-7264577db63f",{"type":24,"contentRole":42,"markdownContent":127,"audioMediaId":128},"We've established that the term 'Artificial Intelligence' can be used to describe a machine or computer which is capable of human-like cognitive processes.\n\nBut there's something we need to clarify: these human-like processes are not the same as genuine human cognition.\n\nIn theory, it's possible that an AI could be built which thinks exactly like a human. An AI whose artificial cognitive processes are indistinguishable from our own.\n\nThere's a name for this: **Artificial General Intelligence (AGI)**. Just imagine a computer with free-flowing thoughts like the ones that you're having right now.","920b4d8a-97a0-4529-b33d-905a652fa604",{"id":130,"data":131,"type":24,"maxContentLevel":27,"version":24,"reviews":134},"96a38e8e-6126-494d-a40f-23a775472ca1",{"type":24,"contentRole":42,"markdownContent":132,"audioMediaId":133},"As things stand, the idea of AGI is purely hypothetical. No one has ever managed to build an Artificial Intelligence that matches (or even comes close) to a human brain.\n\nPlenty of people have *imagined* it. HAL 9000, in Stanley Kubrick's *2001: A Space Odyssey*, is one great example of an Artificial General Intelligence. There are the robots in *Terminator*, *Bladerunner* and *The Matrix*. But outside the realms of science fiction, there's nothing close to it yet.\n\n![Graph](image://554e39d2-2a4e-491c-93c1-7372ceb1b45d \"HAL 9000. Cropped (cropped) by Mandruss (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")\n\nSome people think we'll *never* get close to it. There's certainly no guarantee. Maybe one day we'll manage to build a machine that can laugh, and dream, and fall in love. But for now, that kind of Artificial Intelligence definitely can't be built.","bf767dfa-d551-4711-ab1f-d4ff41e1d999",[135],{"id":136,"data":137,"type":67,"version":24,"maxContentLevel":27},"95b6d38e-6479-4154-8a5b-49749a8da37d",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":138,"binaryCorrect":140,"binaryIncorrect":142},[139],"HAL 9000, in '2001: A Space Odyssey', is a fictional example of what type of AI?",[141],"Artificial General Intelligence (AGI)",[143],"Artificial Narrow Intelligence (ANI)",{"id":145,"data":146,"type":24,"maxContentLevel":27,"version":24,"reviews":149},"1f71853c-1a76-417a-bd31-746e728230a8",{"type":24,"contentRole":42,"markdownContent":147,"audioMediaId":148},"What *can* be built is **Artificial Narrow Intelligence (ANI)**. This term describes an Artificial Intelligence which can simulate a small number of human-like processes, but doesn't come close to the breadth and fluidity of an actual human brain.\n\nIn general, ANI can only perform its human-like processes in the context of a specific task. A task, that is, which it was originally trained and designed for.\n\nChatGPT is a great example. It was designed to generate bodies of text – and it can't really do much else. You couldn't ask it, for example, to play a piece of music, or to predict tomorrow's weather. And it won't suddenly think, \"I want to change my career\", and start studying to become a physician.\n\n![Graph](image://06ccc0cd-f124-4915-aeee-7cc8c7185fe9 \"ChatGPT. Screenshot (Public domain), via Wikimedia Commons\")\n\nAn AGI (like a human) would be adaptable enough to perform tasks that it wasn't built for. But an Artificial Narrow Intelligence can't do this – it's much too simple, and confined to its original functions.","f0cc62a9-7f00-4c37-b84a-51b1cee67c6f",[150],{"id":151,"data":152,"type":67,"version":24,"maxContentLevel":27},"a8584b4d-0d08-435e-965d-58e0a140e48c",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":153,"multiChoiceCorrect":155,"multiChoiceIncorrect":156,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[154],"Which of these would be adaptable enough to perform tasks that it wasn't built for?",[141],[143,157,158],"Both of these","Neither of these",{"id":160,"data":161,"type":24,"maxContentLevel":27,"version":24,"reviews":164},"24a64c78-ff99-4060-8123-68c5d43f19d6",{"type":24,"contentRole":42,"markdownContent":162,"audioMediaId":163},"It can be helpful to think of those two types of AI – Narrow and General – in terms of a ladder. That ladder is a hierarchy of complexity, moving upwards from simple to advanced.\n\nArtificial Narrow Intelligence is sitting on the bottom rung. That's any AI whose cognitive abilities are **weaker** than those of a human. Artificial General Intelligence is on the next rung up. That's any AI with cognitive abilities that are **the same** as those of a human.\n\nLast but not least, there's also Artificial Super Intelligence (ASI), which we haven't had a chance to mention. That's the next rung up above AGI – it's used to describe an AI model with cognitive abilities that are **greater** than those of a human.\n\n![Graph](image://5a3a8392-f75a-4cfe-ab5c-a632eb2b9f24 \"Ladder of Artificial Intelligence.\")\n\nJust like AGI, Artificial Super Intelligence is purely hypothetical. As things stand, we've only ever managed to reach the bottom rung of that ladder.","de2fa5a3-44e0-41d9-9486-0835eedd7bef",[165,174],{"id":166,"data":167,"type":67,"version":24,"maxContentLevel":27},"17305446-1c79-4611-86bb-eb56bda620f9",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":168,"multiChoiceCorrect":170,"multiChoiceIncorrect":171,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[169],"Which of these different types of AI have been built by humans so far?",[143],[141,172,173],"Artificial Super Intelligence (ASI)","All of these",{"id":175,"data":176,"type":67,"version":24,"maxContentLevel":27},"1c8192c3-5267-4edd-a958-4d797a4572c7",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":178,"matchPairsPairs":180,"matchPairsShowExamples":6},6,[179],"Match the pairs below:",[181,184,187,190],{"left":182,"right":183,"direction":27},"Artificial Narrow Intelligence","Cognitive abilities weaker than a human",{"left":185,"right":186,"direction":27},"Artificial General Intelligence","Cognitive abilities equal to a human",{"left":188,"right":189,"direction":27},"Artificial Super Intelligence","Cognitive abilities greater than a human",{"left":191,"right":192,"direction":27},"Artificial Strong Intelligence","Not a type of AI",{"id":194,"data":195,"type":28,"maxContentLevel":27,"version":37,"orbs":198},"83f3ef80-7c8d-4ba0-a6c5-75cdf749817f",{"type":28,"title":196,"tagline":197},"The Rise of AI","From Babbage to ChatGPT",[199,305,420,552],{"id":200,"data":201,"type":42,"version":42,"maxContentLevel":27,"summaryPage":203,"introPage":210,"pages":216},"df7321e5-fe61-4052-bd02-0e07343a6444",{"type":42,"title":202},"AI origins",{"id":204,"data":205,"type":27,"maxContentLevel":27,"version":24},"36ecf303-1af9-45b3-a200-544aed04c734",{"type":27,"summary":206},[207,208,209],"The analytical engine, invented by Charles Babbage, was the world's first computer","Ada Lovelace observed that this engine was not capable of human-like thought","Later, Alan Turing argued that machines in the future would be capable of human-like thought",{"id":211,"data":212,"type":29,"maxContentLevel":27,"version":24},"55ead02f-241b-43e1-a4db-e45a80e70244",{"type":29,"intro":213},[214,215],"What was the world's first computer?","Who is considered the father of modern AI?",[217,230,243,260],{"id":218,"data":219,"type":24,"maxContentLevel":27,"version":24,"reviews":222},"00469471-19ed-4033-9381-a122e3412c34",{"type":24,"contentRole":42,"markdownContent":220,"audioMediaId":221},"For most of this pathway, you'll be learning about the technical details of AI. But before we get into all that, we'd like to set the scene with a little bit of AI history. No, we're not talking about Talos this time. Instead, we're jumping back to a machine called the **analytical engine**.\n\n![Graph](image://e4cf5b2f-2ae9-4f41-8632-aa9ecb3830bb \"Analytical Engine. Label QS:Len,\\\"Babbage's Analytical Engine\\\" by Charles Babbage (CC BY-SA 2.0) \u003Chttps://creativecommons.org/licenses/by-sa/2.0>, via Wikimedia Commons\")\n\nThe analytical engine was the world's first computer. It was invented by Charles Babbage, an English engineer, towards the start of the 1800s. You could feed it punched cards, which functioned like programmes, and it would respond with a printed answer.\n\nAt least, that's what was meant to happen. In the end, Charles Babbage ran out of funding, and never got to finish the project. But the theory behind it was solid. And it got people wondering about something: if the engine had actually been finished, would it have counted as 'intelligent', or not?","bd3afb69-b574-4eec-8e1f-b0bb581a726e",[223],{"id":224,"data":225,"type":67,"version":24,"maxContentLevel":27},"c2096211-69b7-4eb2-99f6-e7ac913d90e4",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":226,"activeRecallAnswers":228},[227],"What was the name of the world's first computer?",[229],"Analytical Engine",{"id":231,"data":232,"type":24,"maxContentLevel":27,"version":24,"reviews":235},"89ee0940-a820-4594-b855-26f7486928bb",{"type":24,"contentRole":42,"markdownContent":233,"audioMediaId":234},"While Babbage was working on his analytical engine, he was supported by **Augusta Ada King**: mathematician, writer, computer programmer, and respectable Countess of Lovelace.\n\nAda Lovelace (as she's often known) wrote extensive notes about the analytical engine's capabilities. And in 1843, she made an important observation: \"the analytical engine has no pretensions whatsoever to originate anything. It can do whatever we know how to order it to perform.\"\n\nIn other words, she was touching upon that modern distinction between computing and artificial intelligence. The analytical engine wasn't 'intelligent', because it could only follow pre-programmed instructions, as opposed to taking the human-like step of 'originating' something new.\n\n![Graph](image://e4e4647a-666b-4d99-ae2d-ca58bf6de21f \"Portrait of Ada King, Countess of Lovelace. (Public domain), via Wikimedia Commons\")","89e87643-8ff3-44c9-8e0c-bb0f2eca3008",[236],{"id":237,"data":238,"type":67,"version":24,"maxContentLevel":27},"d32621e2-d515-43d3-b72c-8db097694057",{"type":67,"reviewType":37,"spacingBehaviour":24,"clozeQuestion":239,"clozeWords":241},[240],"1843, Ada Lovelace observed: \"the analytical engine has no pretensions whatsoever to originate anything.\"",[242],"originate",{"id":244,"data":245,"type":24,"maxContentLevel":27,"version":24,"reviews":248},"b8d94c28-feb9-4135-bc78-5d6f5511b729",{"type":24,"contentRole":42,"markdownContent":246,"audioMediaId":247},"Almost a hundred years after Ada Lovelace, a new figure pushed to the forefront of computing, and picked up the question of artificial intelligence. His name was **Alan Turing** – in a lot of ways, we might think of him as the father of modern AI.\n\n![Graph](image://3bbc0879-38da-4e80-8d15-534841dcf4cc \"Alan Turing (1912-1954) in 1936 at Princeton University (b&w) (Public domain), via Wikimedia Commons\")\n\nIn 1950, he published a paper titled *Computing Machinery and Intelligence*. In this paper, he wanted to consider the question: are machines capable of thought?\n\nYes, said Turing. In theory, a machine is capable of human-like thought. That technology was still a long way off, but one day, thought Turing, humanity would manage to build an intelligent machine.","bb273777-0022-4aad-b87b-b7944d7cfc7d",[249],{"id":250,"data":251,"type":67,"version":24,"maxContentLevel":27},"3e16ac7a-2106-4c21-9c1a-cc7501c9dc78",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":252,"multiChoiceCorrect":254,"multiChoiceIncorrect":256,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[253],"In 1950, Alan Turing wondered if machines were capable of thought. What was his conclusion?",[255],"Yes, but the technology was a long way off",[257,258,259],"No, machines would never be capable of thought","Yes, and the technology was already available","The question was impossible to answer",{"id":261,"data":262,"type":24,"maxContentLevel":27,"version":42,"reviews":265},"eaf966e0-124e-490c-80d2-aff581f3d5bc",{"type":24,"contentRole":42,"markdownContent":263,"audioMediaId":264},"In that paper, Turing also suggested a way to check if a machine can think. This check became known as the **Turing Test**. There are a few variations, but one of these tests might look a little something like this.\n\nA human evaluator (C) is told to speak with two participants (A and B) via text. One of these participants is a human; the other is secretly a machine. Afterwards, the evaluator is asked a question: of the two participants, can they tell which one was the machine?\n\n![Graph](image://f2a71601-a499-470f-87b2-e958886ddc71 \"Turing test diagram by Juan Alberto Sánchez Margallo (CC BY 2.5) \u003Chttps://creativecommons.org/licenses/by/2.5>, via Wikimedia Commons\")\n\nIf the evaluator struggles to identify the machine, then that machine must have displayed some level of human-like behavior. And in the eyes of Turing, that human-like behavior is evidence of human-like thought.\n\nAt the time, this was all theoretical. No machine could have passed the test. But Turing's writings were still influential. This was the very first time that AI had been discussed in such a detailed, deliberate way.","d7ca40f5-0835-48ef-aa09-37b750c34f65",[266,277,294],{"id":267,"data":268,"type":67,"version":42,"maxContentLevel":27},"b8b42c8d-2faa-435f-8882-2f98ad057e49",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":269,"multiChoiceCorrect":271,"multiChoiceIncorrect":273,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[270],"In the Turing Test, how is the evaluator meant to tell the difference between human and machine?",[272],"By communicating via text",[274,275,276],"By analyzing their facial expressions","By listening to their voices","By observing their body language",{"id":278,"data":279,"type":67,"version":24,"maxContentLevel":27},"a78c45c0-bcd1-42a2-b811-a00c3e731f86",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":280,"matchPairsPairs":281,"matchPairsShowExamples":6},[179],[282,285,288,291],{"left":283,"right":284,"direction":27},"Charles Babbage","Invented the world's first computer",{"left":286,"right":287,"direction":27},"Ada Lovelace","Argued that computers lack intelligence",{"left":289,"right":290,"direction":27},"Alan Turing","Argued that computers could be intelligent",{"left":292,"right":293,"direction":27},"None of these","Invented the world's first AI",{"id":295,"data":296,"type":67,"version":24,"maxContentLevel":27},"52ab8c70-cd4a-4e83-87fb-351807bfdb15",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":297,"multiChoiceCorrect":299,"multiChoiceIncorrect":301,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[298],"Was the Analytical Engine ever finished?",[300],"No, due to lack of funding",[302,303,304],"No, due to technical difficulties","No, due to stolen parts","Yes, but it was never tested",{"id":306,"data":307,"type":42,"version":42,"maxContentLevel":27,"summaryPage":309,"introPage":316,"pages":322},"b72fd3bd-b4f6-47dc-b9ac-465fc11aca9f",{"type":42,"title":308},"AI golden age",{"id":310,"data":311,"type":27,"maxContentLevel":27,"version":24},"d404c46b-eb6a-4ed8-bd7f-e34903e431fd",{"type":27,"summary":312},[313,314,315],"Logic Theorist (the world's first AI) was designed to solve mathematical problems","Logic Theorist was a big leap forward, and led to the formal establishment of the field of Artificial Intelligence","This marked the start of an AI golden age, with the invention of more models like Eliza (the world's first chatbot) ",{"id":317,"data":318,"type":29,"maxContentLevel":27,"version":24},"9adac52f-3a00-45a3-8368-9dd10cc281d8",{"type":29,"intro":319},[320,321],"What sparked the beginning of the AI golden age?","What was the world's first AI?",[323,359,376,393],{"id":324,"data":325,"type":24,"maxContentLevel":27,"version":24,"reviews":328},"0cc84290-80b3-4772-a431-fcbf2ab7c672",{"type":24,"contentRole":42,"markdownContent":326,"audioMediaId":327},"In 1950, when Alan Turing was writing his paper about machine intelligence, it was mostly theoretical. No AI models had ever been built – but it wouldn't take long for this to change.\n\nIn 1955, a team of American computer scientists collaborated on a cutting-edge project. Its name was Logic Theorist, and it's generally thought of as the world's very first AI.\n\nLogic Theorist was an Artificial Narrow Intelligence (ANI), which was designed to solve mathematical problems and establish proofs for famous theorems. This was logical reasoning in action – superficially, at least, Logic Theorist was performing a human-like cognitive process.","e0157ea2-8415-44b3-bdfc-44beda9f784b",[329,348],{"id":330,"data":331,"type":67,"version":24,"maxContentLevel":27},"37029b16-f9d4-43c1-bac7-30c9382ce308",{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":332,"multiChoiceQuestion":336,"multiChoiceCorrect":338,"multiChoiceIncorrect":340,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":344,"matchPairsPairs":345},[333,334,335],"42661e37-7495-436b-a519-6b976d7ba479","14a4785e-81f2-43b7-b271-ba31052cd5c3","dae72ea9-487b-4523-84d4-fe9d3e7511be",[337],"Which of these is generally thought of as the world's first AI?",[339],"Logic Theorist",[341,342,343],"Eliza","Mark I Perceptron","Deep Blue",[179],[346],{"left":339,"right":347,"direction":27},"World's first AI",{"id":349,"data":350,"type":67,"version":24,"maxContentLevel":27},"6d1c0417-7b4d-4b24-9537-b95dd21d5448",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":351,"multiChoiceCorrect":353,"multiChoiceIncorrect":355,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[352],"What specific task was Logic Theorist programmed to perform?",[354],"Solve mathematical problems",[356,357,358],"Simulate human conversation","Play chess","Translate languages",{"id":360,"data":361,"type":24,"maxContentLevel":27,"version":24,"reviews":364},"7ce1441d-8d18-484f-b502-1a7162a78b18",{"type":24,"contentRole":42,"markdownContent":362,"audioMediaId":363},"Logic Theorist was a big leap forward. And it made people realize something. This emerging field of 'intelligent computers' didn't really have a name.\n\nIn 1956, a group of leading scientists in the United States – including the team who had worked on Logic Theorist – decided to meet up at Dartmouth College, New Hampshire. There, they formally established the field of Artificial Intelligence, and the name has stuck ever since.\n\nAt the Dartmouth Conference, as this event became known, the scientists also came up with some goals for the field. Logic Theorist was just the beginning – they wanted to start building Artificial Intelligences which could use language, self-improve, and think creatively.","f4896e33-89c8-4345-879f-6380fc813afe",[365],{"id":366,"data":367,"type":67,"version":24,"maxContentLevel":27},"6f34f949-038a-41cb-8d21-25d0cfe97c07",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":368,"multiChoiceCorrect":370,"multiChoiceIncorrect":372,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[369],"At which conference, held in 1956, was the emerging field of Artificial Intelligence officially given a name?",[371],"Dartmouth Conference",[373,374,375],"Harvard Conference","Princeton Conference","Stanford Conference",{"id":377,"data":378,"type":24,"maxContentLevel":27,"version":24,"reviews":381},"ed9691da-f512-4066-be96-5ea4e4c5d18c",{"type":24,"contentRole":42,"markdownContent":379,"audioMediaId":380},"The Dartmouth Conference was followed by an exciting couple of decades, which are sometimes referred to as the AI golden age. Inspired by Logic Theorist, more and more scientists started to build AIs.\n\nMost of these AIs were based on an idea called **symbolic programming**. In simple terms, this meant giving a computer a tree of logical rules. The computer would use this tree of rules to simulate 'reasoning', and 'decision making', and other human-like processes.\n\n![Graph](image://90256d62-b5de-42c7-bb66-bf2944ba30fc \"Example of a tree of rules. (CC0) \u003Chttp://creativecommons.org/publicdomain/zero/1.0/deed.en>, via Wikimedia Commons\")\n\nLogic Theorist was based on this approach. Another famous example was Eliza, the world’s first AI chatbot. Eliza used symbolic programming to simulate the dialogue of a psychoanalyst, basically just spotting key words and patterns in pieces of text, then generating relevant responses.","148258e3-f3fb-466b-908c-844795e5909b",[382],{"id":334,"data":383,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":384,"multiChoiceQuestion":385,"multiChoiceCorrect":387,"multiChoiceIncorrect":388,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":389,"matchPairsPairs":390},[333,330,335],[386],"Which of these is generally thought of as the world's first AI chatbot?",[341],[339,342,343],[179],[391],{"left":341,"right":392,"direction":27},"World's first chatbot",{"id":394,"data":395,"type":24,"maxContentLevel":27,"version":42,"reviews":398},"7673c001-1757-43ec-be79-bef213ed1d57",{"type":24,"contentRole":42,"markdownContent":396,"audioMediaId":397},"You can still find versions of the Eliza chatbot online. Here's an example of a chat with her:\n\n![Graph](image://c821ae03-be7c-4922-b5db-413334d046af \"A conversation with Eliza.\")\n\nThis dialogue isn't perfect. But it's convincing enough that some people who used Eliza, in the 1960s, came away with the impression that they were speaking to an actual person. In other words, the Eliza chatbot could have potentially passed the Turing Test.","3703bad9-9974-405c-9ae2-ace33d28a7f8",[399,409],{"id":400,"data":401,"type":67,"version":24,"maxContentLevel":27},"92c55cfc-0523-453c-b0f4-550f359219bd",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":402,"multiChoiceCorrect":404,"multiChoiceIncorrect":405,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[403],"The 1950s and 1960s are often referred to as what?",[308],[406,407,408],"AI spring","AI winter","AI revolution",{"id":410,"data":411,"type":67,"version":24,"maxContentLevel":27},"29298782-22a8-448a-b00b-b03f63d92817",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":412,"multiChoiceCorrect":414,"multiChoiceIncorrect":416,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[413],"During the AI golden age, most models were based on what approach?",[415],"Symbolic programming",[417,418,419],"Neural networks","Genetic algorithms","Reinforcement learning",{"id":421,"data":422,"type":42,"version":27,"maxContentLevel":27,"summaryPage":423,"introPage":430,"pages":436},"c9b60bd6-a2a0-4215-b8e0-c6c3e25a249a",{"type":42,"title":407},{"id":424,"data":425,"type":27,"maxContentLevel":27,"version":24},"9c61f1cf-ac95-4407-9196-3e2017cff901",{"type":27,"summary":426},[427,428,429],"After the AI golden age, scientists struggled to make any meaningful progress","This led to a a period called the AI winter, when interest and funding dried up","In 1996, Deep Blue played chess against Garry Kasparov, and won 2 out of 6 games",{"id":431,"data":432,"type":29,"maxContentLevel":27,"version":24},"4ce4abf0-2eb0-4445-9c40-f74e0e550b72",{"type":29,"intro":433},[434,435],"Why did the AI golden age collapse?","What was Deep Blue, and how did it shock the world?",[437,450,467,482],{"id":438,"data":439,"type":24,"maxContentLevel":27,"version":24,"reviews":442},"0ad9f918-f096-4aa7-a5ee-e364ef20aca7",{"type":24,"contentRole":42,"markdownContent":440,"audioMediaId":441},"As we've already talked about, the 1950s and 60s were a golden age in the history of Artificial Intelligence. These decades saw the birth of the first AI models, not just Logic Theorist and Eliza, but plenty of others too.\n\nAnd it felt like this was only the start. In 1958, the *New York Times* reported that it was a matter of time before an electronic computer would be able to \"walk, talk, see, write, reproduce itself and be conscious of its existence.\"\n\nPeople were excited. People were hyped. Funding flowed in from all directions. But as it turned out... this boom wouldn't last for long.","d2a65b16-070e-46c2-938c-3b32a53e289f",[443],{"id":444,"data":445,"type":67,"version":24,"maxContentLevel":27},"fed6f58c-1529-4978-bb08-15bb1dde0b6c",{"type":67,"reviewType":37,"spacingBehaviour":24,"clozeQuestion":446,"clozeWords":448},[447],"In 1958, the New York Times reported that it was a matter of time before a computer would be \"conscious of its existence.\"",[449],"conscious",{"id":451,"data":452,"type":24,"maxContentLevel":27,"version":24,"reviews":455},"952c7d31-8d95-4872-b3af-01586ab09426",{"type":24,"contentRole":42,"markdownContent":453,"audioMediaId":454},"The problem with early AI models was that they were painfully limited in scope. These were 'narrow' AIs in the strictest sense of that word – and scientists were struggling to build anything more complex or advanced.\n\nIn one famous example, IBM designed an Artificial Intelligence which could translate Russian sentences into English. But it could *only* translate very simple sentences – this AI knew no more than 6 grammatical rules, and 250 words.\n\nOver the next few years, the US government invested almost $20 million into AI translators like this one. But the work never really got anywhere. In 1966, most of this funding was cut.","71655279-7280-4692-ba3f-f7e369688225",[456],{"id":457,"data":458,"type":67,"version":24,"maxContentLevel":27},"feb0be09-c0fb-4a50-8b44-4d51c276beec",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":459,"multiChoiceCorrect":461,"multiChoiceIncorrect":463,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[460],"What was the main problem with the models being built in the AI golden age?",[462],"They were limited in scope",[464,465,466],"They were too complex","They were badly trained","They didn't work",{"id":468,"data":469,"type":24,"maxContentLevel":27,"version":24,"reviews":472},"7d6cc894-6a31-454b-9da7-df67f383d6d7",{"type":24,"contentRole":42,"markdownContent":470,"audioMediaId":471},"By the 1970s, a lot of people were starting to think that AI was nothing but a gimmick. These models didn't have any real-world uses. They were basically just high-tech toys.\n\nScientists still strove to build something useful. But as hard as they tried, they couldn't manage it. Computing power became a major bottleneck – even when they *thought* of more advanced ideas, the technology wasn't there to support them.\n\nAs more and more people lost interest, and more and more funding dried up, the field entered a period of time which is often called the **AI winter**.\n\n![Graph](image://5e8549b2-ef38-4b15-8b25-2033ae4975af \" \")","6b0dd1db-a66a-4c85-92e5-d509276e14aa",[473],{"id":474,"data":475,"type":67,"version":24,"maxContentLevel":27},"25b39fa8-305d-4f1d-8827-256e708cb11d",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":476,"multiChoiceCorrect":478,"multiChoiceIncorrect":479,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[477],"The AI golden age was followed by what, as research into Artificial Intelligence stagnated?",[407],[406,480,481],"AI recession","AI autumn",{"id":483,"data":484,"type":24,"maxContentLevel":27,"version":27,"reviews":487},"1b3c0fda-29f8-4084-ad1d-423deab8995a",{"type":24,"contentRole":42,"markdownContent":485,"audioMediaId":486},"The AI winter continued, on and off, all the way into the early 2000s. Though it has to be said, there were still some pretty exciting moments on the way.\n\nFor example, in the spring of 1996, the current chess GrandMaster, Garry Kasparov, played a series of games against an Artificial Intelligence named **Deep Blue**.\n\n![Graph](image://e9a73af4-b95d-4397-9bd6-32b1b539a4f9 \"Gary Kasparov. Copyright 2007, S.M.S.I., Inc. - Owen Williams, The Kasparov Agency. (CC BY-SA 3.0) \u003Chttp://creativecommons.org/licenses/by-sa/3.0/>, via Wikimedia Commons\")\n\nDeep Blue used symbolic programming to evaluate hundreds of thousands of chess positions in a single second, then decide how to make the best move. It played 6 games against Kasparov, and while it did lose 4 of them, it impressively managed to win 2.\n\nThis was an exciting development for Artificial Intelligence. But again, it was a bit of a gimmick. A chess-playing robot was fun in theory, but just like Logic Theorist and the Eliza chatbot, it didn't really have any useful applications in practice.","3eccc0c5-e7af-4459-8021-9aa0b8d7140b",[488,506,525,536,545],{"id":489,"data":490,"type":67,"version":24,"maxContentLevel":27},"223d5086-2ea3-46ba-9777-723c9e202ef2",{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":491,"multiChoiceQuestion":495,"multiChoiceCorrect":497,"multiChoiceIncorrect":499,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":503,"matchPairsPairs":504},[492,493,494],"03c9194d-374d-4729-8424-6a0b3b4c8d4b","f23209c6-ee2a-4ec9-878f-ceb80bd6ebae","ba8ca3b7-3f63-4310-854b-13d0f5acd9f1",[496],"In simple terms, how would you describe symbolic programming?",[498],"Building AI with a tree of logical rules",[500,501,502],"Using symbols to represent equations","Using equations to represent symbols","Building AI with a web of artificial neurons",[179],[505],{"left":498,"right":415,"direction":27},{"id":507,"data":508,"type":67,"version":42,"maxContentLevel":27},"ebe7001b-03fa-4e1c-a146-185849d6a729",{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":509,"multiChoiceQuestion":513,"multiChoiceCorrect":515,"multiChoiceIncorrect":517,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":521,"matchPairsPairs":522},[510,511,512],"ada192e0-586e-485b-85f7-f1aa3896e776","e97e3df6-3f86-4e78-b7e3-2c0cf09033b4","60c54579-7ff0-4106-8349-2b965cc0f663",[514],"Which Chess Grandmaster played against Deep Blue in 1996?",[516],"Garry Kasparov",[518,519,520],"Bobby Fischer","Anatoly Karpov","Vladimir Kramnik",[179],[523],{"left":516,"right":524,"direction":27},"Played chess against Deep Blue",{"id":526,"data":527,"type":67,"version":24,"maxContentLevel":27},"705d8441-5dc2-439f-b946-6a29356278a6",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":528,"multiChoiceCorrect":530,"multiChoiceIncorrect":532,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[529],"During the AI golden age, IBM made a Russian-to-English translator. But this model was limited to how many words?",[531],"250",[533,534,535],"500","1000","1500",{"id":537,"data":538,"type":67,"version":24,"maxContentLevel":27},"75fba60b-4484-41f0-9437-3aad39bdb228",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":539,"binaryCorrect":541,"binaryIncorrect":543},[540],"How many games did Deep Blue lose to Garry Kasparov in 1996?",[542],"4",[544],"2",{"id":546,"data":547,"type":67,"version":24,"maxContentLevel":27},"89e3cc69-4bcd-4a61-80af-1002635c4fb3",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":548,"binaryCorrect":550,"binaryIncorrect":551},[549],"How many games did Deep Blue win against Garry Kasparov in 1996?",[544],[542],{"id":553,"data":554,"type":42,"version":24,"maxContentLevel":27,"summaryPage":555,"introPage":562,"pages":568},"d2ece187-aec5-4da0-94dd-7df9307e0b9a",{"type":42,"title":406},{"id":556,"data":557,"type":27,"maxContentLevel":27,"version":24},"970843fa-20cf-444d-b98e-038c23e2352a",{"type":27,"summary":558},[559,560,561],"In 2016, AlphaGo achieved a 4-1 win against Go master Lee Sedol in Korea","AlphaGo used a neural network, which was a powerful new type of AI","AlphaGo marked the start of the AI spring – a period we're still living through now",{"id":563,"data":564,"type":29,"maxContentLevel":27,"version":24},"e8cc373a-5050-4c16-bd58-ff83fa74e9e2",{"type":29,"intro":565},[566,567],"What are neural networks, and why are they so important?","What marked the start of the modern AI spring?",[569,590,603,617],{"id":570,"data":571,"type":24,"maxContentLevel":27,"version":24,"reviews":574},"df6c62f4-d211-47ff-91d6-c24cd8769b4d",{"type":24,"contentRole":42,"markdownContent":572,"audioMediaId":573},"In 2016, exactly twenty years after Deep Blue faced off against Gary Kasparov, a research laboratory named Google DeepMind successfully developed an exciting new AI.\n\nThe name of this AI was **AlphaGo** – and it was designed to play the Chinese game of Go. While the world looked on, it went head-to-head against revered Go master, Lee Sedol, in Korea.\n\n![Graph](image://fb1b34d4-f0f5-47ec-9f66-d81810ee0d13 \"AlphaGo playing Go. Image: Axd (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")\n\nThe game of Go is extremely complex, and much harder to play than chess. Because of this, most people predicted a landslide victory for Lee. But instead, to everyone's general amazement, AlphaGo achieved a stunning 4-1 win.","0d2cda62-2f6e-4902-91e9-115d93acf930",[575],{"id":511,"data":576,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":577,"multiChoiceQuestion":578,"multiChoiceCorrect":580,"multiChoiceIncorrect":582,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":586,"matchPairsPairs":587},[510,507,512],[579],"In 2016, AlphaGo shocked the world when it defeated which revered Go master?",[581],"Lee Sedol",[583,584,585],"Ke Jie","Fan Hui","Cho Chikun",[179],[588],{"left":581,"right":589,"direction":27},"Played Go against AlphaGo",{"id":591,"data":592,"type":24,"maxContentLevel":27,"version":24,"reviews":595},"b78c75b9-eb2b-4afb-ab13-431522e9c793",{"type":24,"contentRole":42,"markdownContent":593,"audioMediaId":594},"Now, it's important to understand that AlphaGo and Deep Blue were two different types of AI. As we've already talked about, Deep Blue relied on symbolic programming – a tree of commands and rules.\n\nBut AlphaGo relied on something called a **neural network**. We'll talk more about these a bit later. But in simple terms, a neural network is a web of artificial neurons. These artificial neurons are linked together by thousands of connections, just like a human brain.\n\nAgain, we'll get into the details later. But here's the important part (for now): this type of AI is a lot more advanced, and a lot more powerful, than traditional symbolic programming.","1c8f73e5-3baa-44e1-9d70-83a454c0ccae",[596],{"id":597,"data":598,"type":67,"version":24,"maxContentLevel":27},"582bbece-f935-4a73-b7dd-1ee0d482ed51",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":599,"activeRecallAnswers":601},[600],"Deep Blue was built around symbolic programming. But what approach was used for AlphaGo?",[602],"Neural network",{"id":604,"data":605,"type":24,"maxContentLevel":27,"version":24,"reviews":608},"58ad7eee-d5c2-464b-a18e-150dc11932d9",{"type":24,"contentRole":42,"markdownContent":606,"audioMediaId":607},"For a lot of people, the success of AlphaGo came to symbolize the end of the AI winter. This was the start of an exciting new period often known as the **AI spring**.\n\nHere's the thing. Neural networks weren't a new idea. Like symbolic programming, they'd been around since the 1950s. But it was only now that computing power was advanced enough to properly unlock their potential.\n\nAlong with AlphaGo, neural networks have also been used to build AI models like ChatGPT. It's like the Eliza chatbot, but *significantly* better – it's so good at generating human-like text that millions of people now use it to help with day-to-day writing tasks.","619d7fb3-b973-44c8-9642-0f6733e22c2b",[609],{"id":610,"data":611,"type":67,"version":24,"maxContentLevel":27},"bb7c05c7-7d80-4ae1-9f6d-088bec4d2933",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":612,"multiChoiceCorrect":614,"multiChoiceIncorrect":615,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[613],"The success of AlphaGo is often seen as the start of what era?",[406],[616,407,308],"AI summer",{"id":618,"data":619,"type":24,"maxContentLevel":27,"version":24,"reviews":622},"849a3e24-e72a-4b45-abb8-99c88251d708",{"type":24,"contentRole":42,"markdownContent":620,"audioMediaId":621},"Along with ChatGPT, the AI spring has also seen other exciting leaps forward in the field of Artificial Intelligence. Google, for example, is building self-driving cars, which use specialized sensors to 'look' at their surroundings, and make sure that they're driving safely.\n\n![Graph](image://31aa36c5-f9a3-47af-95e7-b474d039df67 \"Waymo self-driving car side view. Image by Grendelkhan (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")\n\nIn the field of medicine, AI can be used to analyze x-rays, to develop new treatments, to invent new vaccines and drugs. In business and banking, it can be used to interpret vast amounts of data.\n\nAnd don't forget about education! At Kinnu, our researchers are investigating ways to use AI to make the learning experience more adaptive, more high quality, and more accessible.\n\nOne thing's for certain: this field has come an awfully long way since the days of Charles Babbage and Ada Lovelace. Artificial Intelligence is real, and it's here, and the world won't ever be the same.","c9d90008-5a12-4b1c-ba15-3827aaa4e48c",[623,630],{"id":624,"data":625,"type":67,"version":24,"maxContentLevel":27},"eb2cb323-e8f2-423b-a3f8-baadfa8b6dbc",{"type":67,"reviewType":37,"spacingBehaviour":24,"clozeQuestion":626,"clozeWords":628},[627],"In simple terms, a neural network is a web of artificial neurons.",[629],"neurons",{"id":631,"data":632,"type":67,"version":24,"maxContentLevel":27},"28c3974a-6ea2-40e0-b9b4-fd1b617499da",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":633,"matchPairsPairs":635,"matchPairsShowExamples":6},[634],"Which of these models would you associate with which era?",[636,637,638,640],{"left":341,"right":308,"direction":27},{"left":343,"right":407,"direction":27},{"left":639,"right":406,"direction":27},"ChatGPT",{"left":229,"right":292,"direction":27},{"id":642,"data":643,"type":28,"maxContentLevel":27,"version":27,"orbs":646},"06e8fa70-d951-406c-9911-fcd1faf14c1b",{"type":28,"title":644,"tagline":645},"Symbolic Programming","The good old fashioned way",[647,743,849,939],{"id":648,"data":649,"type":42,"version":24,"maxContentLevel":27,"summaryPage":651,"introPage":658,"pages":663},"6806b536-a84a-4d00-a98e-0a1c25e473a9",{"type":42,"title":650},"What is symbolic programming?",{"id":652,"data":653,"type":27,"maxContentLevel":27,"version":24},"2bc4ea82-4eff-46b5-8238-49994e8b5699",{"type":27,"summary":654},[655,656,657],"Symbolic programming involves giving a computer a tree of symbols and logical rules","Symbols are labels that represent something, similar to how humans use words","Symbolic AI can store these symbols in a library known as a knowledge base",{"id":659,"data":660,"type":29,"maxContentLevel":27,"version":24},"800d26b6-8927-43ae-80f1-ddc3168e4300",{"type":29,"intro":661},[650,662],"Why do some people call it Good Old Fashioned AI?",[664,677,702],{"id":665,"data":666,"type":24,"maxContentLevel":27,"version":24,"reviews":669},"8c8a9ae2-83b7-4884-9d88-b431271312be",{"type":24,"contentRole":42,"markdownContent":667,"audioMediaId":668},"Earlier, we saw how **symbolic programming** was at the heart of the AI golden age. In simple terms, this meant giving a computer a tree of instructions, which effectively allowed it to 'make decisions' by following logical rules.\n\nAs you probably remember, symbolic programming is less popular now than it used to be. Most modern models, like AlphaGo and ChatGPT, use a totally different approach. They make their decisions using a web of neurons, which scientists call a neural network.\n\nWe'll learn all about that later. But first, we're going to take a look at symbolic programming in more detail. Yes, neural networks are more popular and more powerful. But in certain contexts, symbolic AI is still a very effective approach.\n\nSome people even refer to it as Good Old Fashioned Artificial Intelligence (GOFAI). That's why we're starting off with it. Good old fashioned stuff first, neural networks later on.","8d27064a-1e29-48e7-8124-3a74a54e8669",[670],{"id":671,"data":672,"type":67,"version":24,"maxContentLevel":27},"f37ae0eb-cdd0-46e5-a1ce-f77347b396a5",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":673,"activeRecallAnswers":675},[674],"Some people refer to symbolic models as GOFAI. But what does GOFAI stand for?",[676],"Good Old Fashioned Artificial Intelligence",{"id":678,"data":679,"type":24,"maxContentLevel":27,"version":24,"reviews":682},"90551f86-170a-4072-bdf6-9ab2967170d0",{"type":24,"contentRole":42,"markdownContent":680,"audioMediaId":681},"First thing's first: you're probably wondering why we call it symbolic programming. Well, it's because this approach uses **symbols** to represent ideas and objects.\n\nWhen we say 'symbols', we're not talking about hieroglyphics. Instead, we're talking about labels. Labels which represent something. It's actually pretty similar to the way that a human uses words.\n\nWhen you say the word \"apple\", you're just using that word as a symbol for a particular fruit. A particular fruit with a particular taste, and a particular color, which grows on a particular tree.\n\nThe whole idea of 'appleness' is summed up by the word \"apple\". In other words, \"apple\" is just a symbol which represents that idea.\n\n![Graph](image://3ec4cd53-e0c6-460a-95ae-db52d4897307 \"Apple by Abhijit Tembhekar (CC BY 2.0) \u003Chttps://creativecommons.org/licenses/by/2.0>, via Wikimedia Commons\")","fd1d22b7-0616-4456-b48f-ac9eafacbf0a",[683],{"id":684,"data":685,"type":67,"version":24,"maxContentLevel":27},"0133be01-b811-4728-b69c-1fef94437099",{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":686,"multiChoiceQuestion":690,"multiChoiceCorrect":692,"multiChoiceIncorrect":694,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":698,"matchPairsPairs":699},[687,688,689],"6d524319-b4e0-487f-abf1-b926a6d71ca9","7a6b611e-6b5f-4bdd-bd95-10c32e397c86","dc19d651-ce4e-4af1-8819-4ac1d8132d15",[691],"In the context of symbolic programming, what are symbols?",[693],"Labels which represent ideas",[695,696,697],"Functions used for computation","Addresses pointing to data","Instructions which use logic",[179],[700],{"left":693,"right":701,"direction":27},"Symbols",{"id":703,"data":704,"type":24,"maxContentLevel":27,"version":24,"reviews":707},"4d9baed5-4f4f-4faf-8dca-306d9ae09577",{"type":24,"contentRole":42,"markdownContent":705,"audioMediaId":706},"A symbolic AI might have access to a library of thousands of symbols, each representing an idea. We call this library a **knowledge base** – and for good old fashioned symbolic AI, it will need to be programmed by hand.\n\nFor example, a programmer might manually add \"apple\", and associate it with properties like \"fruit\", \"red\", and \"grows on trees\". They might also add \"pineapple\", and associate it with \"fruit\", \"yellow\", and \"grows on ground\". Do this for a hundred types of fruit, and you end up with a decent knowledge base.\n\n![Graph](image://3c05fa02-1cb2-4b97-854e-e2edff5bc343 \"Pineapple by Kaweesaesther (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")\n\nAnd the AI can use this knowledge base to 'think'. For example, you could ask it, \"What is a pineapple?\" and it could tell you \"A pineapple is a yellow fruit which typically grows on the ground.\"","42730c2d-fd32-4d42-b932-cd9a439832bc",[708,727,734],{"id":709,"data":710,"type":67,"version":24,"maxContentLevel":27},"d3dd576b-ff35-4f24-9bc6-47aa854381a4",{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":711,"multiChoiceQuestion":715,"multiChoiceCorrect":717,"multiChoiceIncorrect":719,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":723,"matchPairsPairs":724},[712,713,714],"1a45bab4-a167-4a62-8862-c7d62c2bf564","88922b45-c399-41cf-ab79-bc97e3117e49","c28d6c0f-c32f-45a9-b821-c5292636a9d5",[716],"A symbolic AI may have access to a library of symbols. What is this library called?",[718],"Knowledge base",[720,721,722],"Knowledge graph","Knowledge bank","Knowledge pool",[179],[725],{"left":718,"right":726,"direction":27},"Library of symbols",{"id":728,"data":729,"type":67,"version":24,"maxContentLevel":27},"4e8cf364-1ec3-4af7-8266-00af40bb19ef",{"type":67,"reviewType":37,"spacingBehaviour":24,"clozeQuestion":730,"clozeWords":732},[731],"The word \"apple\" is a symbol which represents information like \"fruit\", \"red\", and \"grows on trees\".",[733],"symbol",{"id":735,"data":736,"type":67,"version":24,"maxContentLevel":27},"904e9a7c-87a8-4d0e-a94a-5b16b68f620a",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":737,"binaryCorrect":739,"binaryIncorrect":741},[738],"With good old fashioned AI models, the knowledge base will always be:",[740],"Programmed manually",[742],"Generated automatically",{"id":744,"data":745,"type":42,"version":24,"maxContentLevel":27,"summaryPage":747,"introPage":754,"pages":760},"c82dc050-c93e-4d07-8009-bfdad68d1260",{"type":42,"title":746},"Propositional logic",{"id":748,"data":749,"type":27,"maxContentLevel":27,"version":24},"a8f283be-c66c-4d0f-bc16-7e6e2edb595b",{"type":27,"summary":750},[751,752,753],"Logic is a set of rules which tells an AI how to 'think' about different symbols","Propositional logic is built around statements that can be either true or false","Logical connectives like IF and THEN can link propositions together",{"id":755,"data":756,"type":29,"maxContentLevel":27,"version":24},"e60c0c78-85d9-4fbe-90e3-e6868885c0b5",{"type":29,"intro":757},[758,759],"What is the role of logic in symbolic AI?","How can propositional logic help AI to make decisions?",[761,780,801,817],{"id":762,"data":763,"type":24,"maxContentLevel":27,"version":24,"reviews":766},"60ed5dd6-3ca1-4bfc-9a84-e74b85245736",{"type":24,"contentRole":42,"markdownContent":764,"audioMediaId":765},"So, we know that symbolic AI is built around a bank of pre-programmed symbols. We looked at a fruit example: \"apple\" might be a symbol for a number of properties, including \"red\" and \"grows on trees\".\n\nWhat we now need to talk about is how these symbols relate to **logic**.\n\nLogic, in this context, is just a set of rules which tell an AI how to 'think' about different symbols. Here's an analogy: symbols are the building blocks of symbolic AI, while logic is a detailed instruction guide which tells the AI how to arrange those blocks into shapes.\n\nThat isn't a perfect analogy. But it's a helpful idea to bear in mind as we look at logic in more detail.","369a983f-55bb-44c3-97a6-27ab6f95f03e",[767],{"id":688,"data":768,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":769,"multiChoiceQuestion":770,"multiChoiceCorrect":772,"multiChoiceIncorrect":774,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":776,"matchPairsPairs":777},[684,687,689],[771],"In the context of symbolic programming, what is logic?",[773],"Rules which relate to symbols",[693,696,775],"Variables used for computation",[179],[778],{"left":773,"right":779,"direction":27},"Logic",{"id":781,"data":782,"type":24,"maxContentLevel":27,"version":24,"reviews":785},"dc70e0c0-8640-4481-9274-57c06780fe1f",{"type":24,"contentRole":42,"markdownContent":783,"audioMediaId":784},"There are actually a few different types of logic. We're going to start with the simplest: **propositional logic**.\n\nPropositional logic is built around statements which can either be true or false. For example, a statement like \"apples are edible\" (true) or \"bananas are purple\" (false). We call these statements propositions.\n\nThese propositions will need to be programmed manually. But they will usually contain a symbol that's already in the system. For example, \"the box contains a banana\" is a proposition which includes the symbol \"banana\".\n\n![Graph](image://c6e39705-09fa-4c80-af81-ecdd4d423634 \"Banana box. (Public domain), via Wikimedia Commons\")\n\nA proposition might switch back and forth between 'true' and 'false'. Sometimes, there might be a banana in that box. Sometimes, there might be an apple. But it will always be one or the other: these statements should never be both 'true' and 'false' at once.","d9d2ceac-c4c7-4e59-ac23-f225d8ebbbc5",[786],{"id":689,"data":787,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":788,"multiChoiceQuestion":789,"multiChoiceCorrect":791,"multiChoiceIncorrect":793,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":797,"matchPairsPairs":798},[684,687,688],[790],"Propositional logic is built around propositions. But what are these, exactly?",[792],"True or false statements",[794,795,796],"Rhetorical questions","Logical connectives","Predicate triplets",[179],[799],{"left":792,"right":800,"direction":27},"Propositions",{"id":802,"data":803,"type":24,"maxContentLevel":27,"version":24,"reviews":806},"b1bda9fb-7561-4b7d-b1ae-454cada53e60",{"type":24,"contentRole":42,"markdownContent":804,"audioMediaId":805},"Along with true-or-false propositions, propositional logic also uses **logical connectives**. These are words like IF and THEN, which can be used to link a series of statements together.\n\nLet's say you had two propositions: \"it is raining\" and \"you need an umbrella\". Using logical connectives, you could link those together into \"IF it is raining THEN you need an umbrella\".\n\nIn effect, these connectives turn the true-or-false propositions into a decision tree. And if you tell the AI \"I'm about to go out, should I bring an umbrella?\" it could use this tree to give you a helpful answer.\n\n![Graph](image://cf4c58e1-1229-400d-bcf4-02eda220990d \" \")\n\nIf \"it is raining\" is true, the AI will tell you \"yes, you need an umbrella.\" If \"it is raining\" is false, the AI will tell you \"no, you don't need one today\".","a6ba5fd4-60f2-4323-9b44-b06444f86218",[807],{"id":808,"data":809,"type":67,"version":24,"maxContentLevel":27},"1e6112cc-ba69-4f99-bea8-f6dfc462ec65",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":810,"multiChoiceCorrect":812,"multiChoiceIncorrect":813,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[811],"Words like IF and THEN can be used to link propositions. But what do we call these links?",[795],[814,815,816],"Propositional connectives","Logical statements","Propositional statements",{"id":818,"data":819,"type":24,"maxContentLevel":27,"version":24,"reviews":822},"486bfd78-1df6-4c15-a58e-61155b31b04d",{"type":24,"contentRole":42,"markdownContent":820,"audioMediaId":821},"Along with IF and THEN, two more examples of logical connectives are AND and OR. Again, these connectives can be used to turn a set of true-or-false propositions into a decision tree.\n\nLet's look at another example: \"IF you keep coughing OR you keep sneezing AND you have a temperature THEN you might have the flu\".\n\n![Graph](image://4ed9e780-5d7c-4c0e-a6b5-8ce8285ed08f \" \")\n\nThis is still quite a simple example. But imagine if you programmed a symbolic AI with *every* possible medical symptom, plus *every* possible diagnosis? You'd end up with a pretty impressive model. A computer that uses thousands of symbols, and massive decision trees, to 'think' like an AI doctor.","936a2958-9b5a-45f5-bb78-6da6783f0675",[823,835],{"id":824,"data":825,"type":67,"version":24,"maxContentLevel":27},"4b4a0daa-a397-437f-8925-eb908a9442ff",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":826,"multiChoiceCorrect":828,"multiChoiceIncorrect":832,"multiChoiceMultiSelect":21,"multiChoiceRevealAnswerOption":6},[827],"Which of these are examples of propositions?",[829,830,831],"cheese is edible","penguins can fly","cats are mammals",[833,834],"where is the dog?","close the door",{"id":836,"data":837,"type":67,"version":24,"maxContentLevel":27},"a286042d-78d0-4ae4-a7f9-02f264a014a4",{"type":67,"reviewType":15,"spacingBehaviour":24,"orderAxisType":177,"orderQuestion":838,"orderItems":840},[839],"Complete this string of propositional logic: ",[841,843,845,847],{"label":842,"sortOrder":4},"IF it is raining",{"label":844,"sortOrder":24},"OR it is snowing",{"label":846,"sortOrder":42},"THEN you need a coat",{"label":848,"sortOrder":27},"AND you need an umbrella ",{"id":850,"data":851,"type":42,"version":27,"maxContentLevel":27,"summaryPage":853,"introPage":860,"pages":866},"524fd25b-7e7e-4a94-9535-ff3d70913728",{"type":42,"title":852},"Predicate logic",{"id":854,"data":855,"type":27,"maxContentLevel":27,"version":24},"74aab9dc-ac40-4a25-ad79-76f7492c165e",{"type":27,"summary":856},[857,858,859],"Predicate logic combines symbols and properties into triplets","These triplets can be arranged in complex network of nodes and edges called a knowledge graph","A knowledge graph allows an AI to make inferences based on relationships between different symbols",{"id":861,"data":862,"type":29,"maxContentLevel":27,"version":24},"91d4eea7-74d8-434b-93a3-a12c99b6f724",{"type":29,"intro":863},[864,865],"How is predicate logic different to propositional logic?","How does predicate logic help AI to think like we do?",[867,888,893,906],{"id":868,"data":869,"type":24,"maxContentLevel":27,"version":42,"reviews":872},"2483ac32-fa00-466c-8f7d-4e7bd24b45a7",{"type":24,"contentRole":42,"markdownContent":870,"audioMediaId":871},"Along with propositional logic, a symbolic AI might also use another type of logic, which is known as **predicate logic**.\n\nWith propositional logic, we saw how a symbol and its properties could be turned into true-or-false propositions. For example, \"apple\" (symbol) and \"can be red\" (property) could be turned into \"apples can be red (true)\".\n\nPredicate logic follows a similar principle, but instead of combining the symbol and the property into a statement, it combines them into something called a triplet. This triplet consists of three parts: subject, predicate and object.\n\nIn the example above, \"apple\" is the subject, \"can be\" is the predicate, and \"red\" is the object. In another example, \"dog\" might be the subject, \"is\" might be the predicate, \"mammal\" might be the object.\n\n![Graph](image://5d471ec9-ae81-46c4-b3e1-091009278c96 \"Dog. Image via Pexels\")","e1ae52eb-0e34-4e45-9ef0-ca76d8b33591",[873],{"id":874,"data":875,"type":67,"version":42,"maxContentLevel":27},"8cf457c4-1dc2-4d55-85b0-67151e6f5ea9",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":876,"matchPairsPairs":878,"matchPairsShowExamples":6},[877],"Match the pairs of this triplet: \"cat is chasing mouse\"",[879,882,885],{"left":880,"right":881,"direction":27},"cat","Subject",{"left":883,"right":884,"direction":27},"is chasing","Predicate",{"left":886,"right":887,"direction":27},"mouse","Object",{"id":889,"data":890,"type":24,"maxContentLevel":27,"version":24},"e510e5ff-ba5e-46a8-aecb-0ff33c993077",{"type":24,"contentRole":42,"markdownContent":891,"audioMediaId":892},"Predicate logic lets symbolic AI do some pretty interesting things.\n\nImagine two symbols. The first symbol, \"dog\", has \"is mammal\" attached to it. The second symbol, \"mammal\", has \"is warm blooded\" attached to it.\n\nYour AI turns these examples into triplets: \"dog\" (subject) \"is\" (predicate) \"mammal\" (object), and \"mammal\" (subject) \"is\" (predicate), \"warm blooded\" (object).\n\nLet's present those two triplets in a table, like the one below.\n\n![Graph](image://5805f781-098f-4014-bbc6-fc288c03bb1a \" \")","b2d8c8a7-5ac5-4a61-b1bd-ea056e31974f",{"id":894,"data":895,"type":24,"maxContentLevel":27,"version":42,"reviews":898},"d7a66fe8-1d6c-495d-bd30-f70794d29a22",{"type":24,"contentRole":42,"markdownContent":896,"audioMediaId":897},"So what's the point of these triplets? Let's take a look.\n\nImagine that you wanted to ask your AI whether dogs are warm blooded. Unfortunately, \"is warm blooded\" wasn't one of the properties attached to the symbol \"dog\". In other words, the AI doesn't know the answer.\n\nHowever... using some predicate logic, this symbolic AI might be able to work something out.\n\nIt looks at those triplets, and identifies that mammals are warm blooded. It *also* identifies that dogs are mammals. And if dogs are mammals, and mammals are warm blooded...\n\n![Graph](image://8d72c21e-c67b-426d-ade6-f34d463de143 \" \")\n\n\"Yes,\" the AI tells you. \"I believe that dogs are warm blooded.\"","6406e2dd-d640-4acf-971b-52279d6979b4",[899],{"id":900,"data":901,"type":67,"version":24,"maxContentLevel":27},"85dcab8e-5d44-4420-a307-3f9777f9d459",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":902,"activeRecallAnswers":904},[903],"Using predicate knowledge, what could an AI deduce from the following triplets? \"ducks are birds\" and \"birds have feathers\".",[905],"Ducks have feathers",{"id":907,"data":908,"type":24,"maxContentLevel":27,"version":24,"reviews":911},"785c0f05-bdd5-459c-9b36-7aa2e2faf69e",{"type":24,"contentRole":42,"markdownContent":909,"audioMediaId":910},"Just like propositional logic, predicate logic can be a very powerful tool.\n\nIn some symbolic AI models, information is arranged into something called a **knowledge graph**. This is a complex network of nodes and edges, with the nodes representing subjects and objects, and the edges representing the predicate relationships between them.\n\n![Graph](image://e2c425e9-2c63-4a4d-a709-ae5e134ddf2a \"Example of a knowledge graph. By Fuzheado (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")\n\nA knowledge graph can help a symbolic AI to draw sophisticated, human-like deductions. It can wind along the graph from node to node, drawing long strings of logic from subject to object to subject to object until it reaches new decisions and conclusions.\n\nUser: \"Could Artificial Intelligence exist without sunlight?\"\n\nAI: \"AI is made by humans, and humans eat animals, and animals eat plants, and humans also eat plants, and plants need sunlight, so no, in conclusion, Artificial Intelligence could not exist without sunlight.\"\n\n![Graph](image://fdabdf3d-d78a-4f4f-b77f-b8335db27f7f \" \")","0ef69887-805f-4573-bec6-f8332e4f7f73",[912,923,932],{"id":712,"data":913,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":914,"multiChoiceQuestion":915,"multiChoiceCorrect":917,"multiChoiceIncorrect":918,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":919,"matchPairsPairs":920},[709,713,714],[916],"Predicate triplets can be arranged in a network of nodes and edges known as what?",[720],[718,721,722],[179],[921],{"left":720,"right":922,"direction":27},"Network of predicate triplets",{"id":924,"data":925,"type":67,"version":24,"maxContentLevel":27},"3bde3242-3a7b-466c-9f70-24fff30e6e44",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":926,"binaryCorrect":928,"binaryIncorrect":930},[927],"In a knowledge graph, what do nodes represent?",[929],"Subjects and objects",[931],"Predicate relationships",{"id":933,"data":934,"type":67,"version":24,"maxContentLevel":27},"5e4bc0a7-dd57-416c-af10-984faf4bb2f6",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":935,"binaryCorrect":937,"binaryIncorrect":938},[936],"In a knowledge graph, what do edges represent?",[931],[929],{"id":940,"data":941,"type":42,"version":42,"maxContentLevel":27,"summaryPage":943,"introPage":950,"pages":956},"68f26d51-95fa-4cfe-abd4-301ee1302d16",{"type":42,"title":942},"Types of knowledge",{"id":944,"data":945,"type":27,"maxContentLevel":27,"version":24},"84238b2c-a30c-41c1-a038-5ce6223c60bd",{"type":27,"summary":946},[947,948,949],"Expert systems were designed to mimic the decision-making skills of human experts","They stored different types of knowledge, like declarative, procedural, and heuristic ","Expert systems have fallen out of fashion, mainly because they're so hard to manually programme",{"id":951,"data":952,"type":29,"maxContentLevel":27,"version":24},"8a3c4227-4719-4a0f-93a6-4f17bb2f9f54",{"type":29,"intro":953},[954,955],"What are expert systems, and how do they work?","What are some of the different types of knowledge?",[957,970,975,997],{"id":958,"data":959,"type":24,"maxContentLevel":27,"version":42,"reviews":962},"0c11f9ef-06ca-42aa-8661-ca53fc57f904",{"type":24,"contentRole":42,"markdownContent":960,"audioMediaId":961},"In the 1960s, symbolic AI was used to build an exciting new type of machine.\n\nIt was known as an **expert system**, and it was designed to mimic the decision-making skills of human experts, like doctors, lawyers, and financial advisors.\n\nThese expert systems had two main parts. First, a colossal knowledge base, holding thousands of relevant symbols. For example, a medical expert system would have a knowledge base of symptoms and diseases, while a legal expert system would have a knowledge base of laws and case studies.\n\nThe second part was called the inference engine: a piece of software which applied propositional logic and predicate logic to the giant knowledge base. This logic allowed the system to 'make decisions'.\n\n![Graph](image://351f6946-5e80-4751-a368-b4aaa2a23e8c \"Parts of an expert system.\")","f6dfb444-b8be-424f-85b2-046ede4eccb5",[963],{"id":964,"data":965,"type":67,"version":24,"maxContentLevel":27},"184bb25e-13f3-45d4-8427-f20dffe3a609",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":966,"activeRecallAnswers":968},[967],"What was the primary purpose of an expert system?",[969],"To mimic the decision-making skills of human experts",{"id":971,"data":972,"type":24,"maxContentLevel":27,"version":24},"7d069ee5-3638-4ee3-8608-b62b55bd13ae",{"type":24,"contentRole":42,"markdownContent":973,"audioMediaId":974},"The knowledge base in an expert system could hold many different types of knowledge. If you don't know what we mean by 'types of knowledge'... well, let's look at a few examples.\n\nFirst, we have something called **declarative knowledge**. This is basically just a simple, solid fact. You might have the symbol, \"coughing\", with \"is a symptom of flu\". Or the symbol \"theft\", with \"is illegal\".\n\nSecond, we have **procedural knowledge**. This is more like a process, or a series of steps. For example, instead of saying that \"coughing is a symptom of flu\", procedural knowledge might be step-by-step instructions that explain how to treat this illness.","54dac688-f9f4-4415-9018-08e4e912a776",{"id":976,"data":977,"type":24,"maxContentLevel":27,"version":24,"reviews":980},"056591b5-e10c-4f5b-ab79-d09be0f1eb02",{"type":24,"contentRole":42,"markdownContent":978,"audioMediaId":979},"Then there's **heuristic knowledge**, also known as a 'rule of thumb'. This describes general guidelines, or approximate strategies, which might help the AI take shortcuts towards better decisions.\n\nFor example, here's a useful 'rule of thumb' for a medical system: \"if symptoms persist, it's worth seeing an actual doctor.\" Or here’s one for a legal system: “if this is a first offense, the punishment should be more lenient.”\n\nHumans will often use 'rules of thumb' when they're making real-life decisions. Heuristic knowledge is just a way to allow symbolic AI to do the same.\n\n![Graph](image://27a9d393-01cf-4eca-a764-944631981568 \"Thumb. Image via Pexels\")","09e186e4-28e8-4be0-a992-cbf18ab1cb95",[981],{"id":982,"data":983,"type":67,"version":24,"maxContentLevel":27},"afd67897-2bd9-4e86-8bed-44fbfb659e27",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":984,"matchPairsPairs":985,"matchPairsShowExamples":6},[179],[986,989,992,995],{"left":987,"right":988,"direction":27},"Declarative knowledge","Simple facts",{"left":990,"right":991,"direction":27},"Procedural knowledge","Series of steps",{"left":993,"right":994,"direction":27},"Heuristic knowledge","Rules of thumb",{"left":292,"right":996,"direction":27},"General impressions",{"id":998,"data":999,"type":24,"maxContentLevel":27,"version":24,"reviews":1002},"4a4d2057-6d6f-412d-aa9d-c9ea64d72216",{"type":24,"contentRole":42,"markdownContent":1000,"audioMediaId":1001},"When expert systems were first introduced, there was a lot of hype around them. By the 1980s, they were being used by many of the world's top businesses, and even some universities.\n\nAs we've already seen, this hype died away during the AI winter. At the end of the day, symbolic AI was too reliant on manual programming. At higher levels of complexity, there were simply too many symbols and statements for anyone to feasibly produce.\n\nBut despite that fact, there's still a place for symbolic AI today. Many doctors, for example, use modern equivalents of expert systems to help them diagnose illnesses. Plenty of businesses and research labs use knowledge graphs to organize data. Even robots and drones use symbolic logic: IF battery is low, THEN it is time to recharge.\n\nThink of it like this: with symbolic AI, there's a threshold of complexity that manual programming can't cross. But for tasks that fall below this threshold? Symbolic AI can still be an effective approach.","7673bdd6-9566-4f55-874c-9591a2d8dec6",[1003,1012,1023],{"id":1004,"data":1005,"type":67,"version":24,"maxContentLevel":27},"c60566a3-1aa0-4741-aec3-d9c3755d839a",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":1006,"binaryCorrect":1008,"binaryIncorrect":1010},[1007],"What is the role of the inference engine in an expert system?",[1009],"It applies logic to the knowledge base",[1011],"It constructs the knowledge base",{"id":1013,"data":1014,"type":67,"version":24,"maxContentLevel":27},"2bad8ffb-225d-4e70-831c-707bcfa3eb69",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":1015,"multiChoiceCorrect":1017,"multiChoiceIncorrect":1019,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[1016],"What are the two main parts of an expert system?",[1018],"Knowledge base & inference engine",[1020,1021,1022],"Data base & inference engine","Data base & decision engine","Knowledge base & decision engine",{"id":1024,"data":1025,"type":67,"version":24,"maxContentLevel":27},"40f181cc-bd55-4fc3-a294-f351f0310964",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":1026,"activeRecallAnswers":1028},[1027],"Nowadays, why is symbolic programming only really used for simple AI models?",[1029],"Because manual programming of complex models is too difficult",{"id":1031,"data":1032,"type":28,"maxContentLevel":27,"version":42,"orbs":1035},"9aa45ec4-8224-42b9-a187-4dee7e236a4f",{"type":28,"title":1033,"tagline":1034},"Machine Learning","Training like a brain",[1036,1095,1164,1258,1355,1446,1543],{"id":1037,"data":1038,"type":42,"version":24,"maxContentLevel":27,"summaryPage":1040,"introPage":1047,"pages":1053},"efeebd88-df18-427e-a761-59c3b12ebbf5",{"type":42,"title":1039},"What is machine learning?",{"id":1041,"data":1042,"type":27,"maxContentLevel":27,"version":24},"7bcecddc-bbad-4f65-b366-e6c9828cb880",{"type":27,"summary":1043},[1044,1045,1046],"Machine learning is an approach to AI models which mimics human learning","An AI is said to 'learn' when it changing its behavior based on an input","Machine learning is often faster and more effective than traditional symbolic programming",{"id":1048,"data":1049,"type":29,"maxContentLevel":27,"version":24},"2126018d-eed4-42c9-875f-2e9e347043bc",{"type":29,"intro":1050},[1051,1052],"What is the alternative to manual programming?","How did this alternative change the world?",[1054,1059,1072],{"id":1055,"data":1056,"type":24,"maxContentLevel":27,"version":24},"b7466057-3b69-49b4-9f70-b8047e8c1ac8",{"type":24,"contentRole":42,"markdownContent":1057,"audioMediaId":1058},"Last time, we saw how symbolic AI has a major limitation. It relies on manual programming – and manual programming can only take us so far. In order to build more complex AIs, we need to use a different approach.\n\nBut what's the alternative to manual programming?\n\nTo answer that question, we can look at a human brain. Your own brain (obviously) wasn't manually programmed. Instead, you filled it with facts and ideas via a process of gradual learning. Taste an apple? That taste was added to your brain. Ride a bike? That skill was automatically stored.\n\nOver the last few decades, many scientists have taken the same approach to AI. They call this approach **machine learning**, and it's literally changed the world.","36eca252-558b-4f4a-aee3-e92f7dd6106c",{"id":1060,"data":1061,"type":24,"maxContentLevel":27,"version":24,"reviews":1064},"33e34d80-c1b6-46af-9c10-e1e712f4652e",{"type":24,"contentRole":42,"markdownContent":1062,"audioMediaId":1063},"We'll get into the complexities of machine learning later, but here's a simple definition. An AI is said to 'learn' when you give it an input – like a big list of numbers, or some photos, or videos – and it changes its behavior afterwards. As we said, this is how humans learn as well. We experience an input, then we change.\n\nLet's run with that photo example. Imagine you've built an Artificial Intelligence that struggles to tell the difference between cats and dogs. You decide to show it a pile of animal photos and tell it over and over: \"this one’s a cat, this one’s a cat, this one’s a dog, this one’s a dog...\"\n\n![Graph](image://45419589-20de-4180-8724-3e3f8ab53ff2 \"Cats and dogs.\")\n\nYour AI starts to notice patterns in the photos, like the shape of the ears or the whiskers. And by the end, it's pretty good at recognizing cats and dogs on its own.","e21cdc57-0223-4607-a6a0-29876cb2f612",[1065],{"id":1066,"data":1067,"type":67,"version":24,"maxContentLevel":27},"a7672cee-1b2a-4c12-8ad4-549f4641bd46",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":1068,"activeRecallAnswers":1070},[1069],"What do we call it when a computer receives an input, then changes its behavior?",[1071],"Machine learning",{"id":1073,"data":1074,"type":24,"maxContentLevel":27,"version":24,"reviews":1077},"e0961c14-1ff2-4be0-83c1-d96473a68c3a",{"type":24,"contentRole":42,"markdownContent":1075,"audioMediaId":1076},"Learning wouldn't be the only way to get an AI to recognize cats. Alternatively, a scientist could write symbolic rules (\"fluffy tail\", \"pointy ears\", \"long whiskers\", etc.), then manually build these rules into the AI's design.\n\nBut it's often a lot faster, and a lot more effective, to take a learning approach instead.\n\nLet's think about another example: an AI that's designed to translate text. Instead of manually programming all the words from every language, you can train the AI on a pile of texts which have already been translated. As it works through those texts, and compares the languages, it can learn all the different words.\n\nThis is exactly what Google did in 2016. They built an AI, then trained it on millions of translated texts. Whenever you load up Google Translate, you're interacting with an impressive, self-taught AI.\n\n![Graph](image://cd3c914a-7b0d-41a2-82eb-1a294c3cfeb8 \"Google Translate. Image (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")","aa7c5ab6-415f-4623-8abb-cf0f9bab7781",[1078,1087],{"id":1079,"data":1080,"type":67,"version":24,"maxContentLevel":27},"a633e54d-a520-4349-baab-de0c43fda69b",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":1081,"binaryCorrect":1083,"binaryIncorrect":1085},[1082],"In 2016, how did Google build the AI model which powers Google Translate?",[1084],"They trained it on millions of translated texts",[1086],"They programmed it with the help of native speakers",{"id":1088,"data":1089,"type":67,"version":24,"maxContentLevel":27},"8305fa05-379a-4555-8af3-b246823b028f",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":1090,"binaryCorrect":1092,"binaryIncorrect":1093},[1091],"When you want an AI to perform a task, which approach is generally faster?",[1071],[1094],"Manual programming",{"id":1096,"data":1097,"type":42,"version":24,"maxContentLevel":27,"summaryPage":1099,"introPage":1106,"pages":1112},"8094f498-508a-47da-8d93-835aed730456",{"type":42,"title":1098},"Generalization",{"id":1100,"data":1101,"type":27,"maxContentLevel":27,"version":24},"bec272d3-6246-4006-823a-303718d5e862",{"type":27,"summary":1102},[1103,1104,1105],"A model's ability to handle data it hasn't seen before is known as generalization","Manually programmed AI models typically struggle with generalization","Machine learning models, on the other hand, are good at generalization",{"id":1107,"data":1108,"type":29,"maxContentLevel":27,"version":24},"52d7a44d-af94-4e12-a96f-0eabf2eaeb2c",{"type":29,"intro":1109},[1110,1111],"What is generalization, and why is it so important?","Are AI models capable of generalization or not?",[1113,1118,1131],{"id":1114,"data":1115,"type":24,"maxContentLevel":27,"version":24},"7ace9664-80bf-4524-a3aa-dab28a6ea170",{"type":24,"contentRole":42,"markdownContent":1116,"audioMediaId":1117},"So, we know that machine learning will often be faster and easier than manual programming. But there's also another important benefit. If it's done correctly, this approach can actually allow a model to 'think outside the box'.\n\nImagine that you manually programmed an AI to close your window when it’s rainy, and to open your window when it’s sunny. Then one day it snows, and the AI doesn’t know what to do.\n\nIt's another big drawback with symbolic AI. However much time you spend on the programming, you might struggle to think of every possible scenario. And if you forget a scenario – if you don't programme a rule – your AI model won't know how to respond.\n\n![Graph](image://13600504-58cb-43d8-8df4-a8cd12a12a2c \"Snow. Image via Pexels\")","4e13d5e2-408f-4228-a5b4-2a688f7f829e",{"id":1119,"data":1120,"type":24,"maxContentLevel":27,"version":24,"reviews":1123},"20602a37-9529-4b55-a555-d8ea261ed2e2",{"type":24,"contentRole":42,"markdownContent":1121,"audioMediaId":1122},"Compare that to machine learning, where an AI model has watched thousands of videos of people opening and closing their windows.\n\nNone of these videos have snow in them – but the AI has noticed some patterns. People usually close their windows when something falls from the sky. Especially when that something is cold and wet.\n\nBased on these patterns, the AI is able to extrapolate: the snow appears to be cold and wet, so it must be time to close the window.","ab12f439-5b21-4bc6-b607-629feb63ca2c",[1124],{"id":1125,"data":1126,"type":67,"version":24,"maxContentLevel":27},"03c6b68d-6d2f-4294-bb46-2d0f71dffbde",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":1127,"binaryCorrect":1129,"binaryIncorrect":1130},[1128],"Which approach is better for preparing a model to deal with unexpected scenarios?",[1071],[1094],{"id":1132,"data":1133,"type":24,"maxContentLevel":27,"version":24,"reviews":1136},"dbba8a52-9411-4e55-a8a1-7b459b2d61bf",{"type":24,"contentRole":42,"markdownContent":1134,"audioMediaId":1135},"That snow scenario is an example of **generalization**. In the context of AI, this term refers to a model's ability to handle data it hasn't seen before, as opposed to just handling the data it originally learned from.\n\nHumans are pretty good at this. Imagine showing a child some pictures of dogs: big dogs, small dogs, furry dogs, short-haired dogs. Later, in the park, they see a breed of dog that wasn't in the photos. But they still know it's a dog, because they have a 'generalized' understanding of the data.\n\n![Graph](image://00510760-a931-406e-a9e4-b5a4b162fb17 \"Image via Pexels\")\n\nMachine learning allows an AI model to gain a generalized understanding too. This is extremely useful – as we said, it's hard to predict (and manually programme) every possible scenario that an AI model might face.","8b1760d8-4872-41f9-8ef9-1d2b8b764498",[1137,1155],{"id":1138,"data":1139,"type":67,"version":24,"maxContentLevel":27},"4fb75b36-77c9-4875-890f-79bad100dc81",{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":1140,"multiChoiceQuestion":1144,"multiChoiceCorrect":1146,"multiChoiceIncorrect":1147,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":1151,"matchPairsPairs":1152},[1141,1142,1143],"ea05be6e-cdeb-44f6-9f31-8aec11f44810","2e5c3c78-0e9c-4d53-996b-4757282d0c67","40b646f9-87d1-42a7-affa-48dfb7102234",[1145],"What do we call a model's ability to handle data it hasn't seen before?",[1098],[1148,1149,1150],"Overfitting","Specialization","Adaptation",[179],[1153],{"left":1098,"right":1154,"direction":27},"Ability to handle new data",{"id":1156,"data":1157,"type":67,"version":24,"maxContentLevel":27},"1528186c-723c-4a4e-b608-d32b1d5cc56b",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":1158,"binaryCorrect":1160,"binaryIncorrect":1162},[1159],"True or false: humans are generally pretty good at generalization.",[1161],"TRUE",[1163],"FALSE",{"id":1165,"data":1166,"type":42,"version":24,"maxContentLevel":27,"summaryPage":1168,"introPage":1175,"pages":1181},"fbc37b82-52c0-40e6-97b4-bd980149aa8f",{"type":42,"title":1167},"Parameters and loss",{"id":1169,"data":1170,"type":27,"maxContentLevel":27,"version":24},"5c8f3e76-fe4a-4e1a-802e-e6aabc7c7b3d",{"type":27,"summary":1171},[1172,1173,1174],"When a machine is 'learning', it simply means that numerical parameters are changing up and down","Loss function describes the gap between an AI's output, and a real-life piece of data","The smaller the loss function, the better the machine has learned",{"id":1176,"data":1177,"type":29,"maxContentLevel":27,"version":24},"b4298f4e-9655-4d43-80b1-b6aa1ccf56c4",{"type":29,"intro":1178},[1179,1180],"What does it actually mean when we say that a machine is 'learning'?","Why do larger datasets make machine learning more effective?",[1182,1203,1208,1221],{"id":1183,"data":1184,"type":24,"maxContentLevel":27,"version":24,"reviews":1187},"bd6960df-d290-4801-b7dc-1fa3ae9ce945",{"type":24,"contentRole":42,"markdownContent":1185,"audioMediaId":1186},"Like everything in AI, machine learning is essentially a computer programme which mimics a human-like process. But how does this programme work?\n\nYou can think of it as a template with some numbers attached. AI scientists refer to these numbers as **parameters**.\n\n![Graph](image://61864a07-3264-42a2-b193-74302e343166 \"A (very) simple template with parameters.\")\n\nWhen data is fed into this template, those parameters automatically adjust. The numbers will either get higher or lower, depending on the nature of the data. Just imagine someone sitting there, adjusting dials up and down, as the different bits of data come through.\n\nWe'll look at an example in a second. But in simple terms: when we say that a machine is 'learning', we really just mean that these parameters are changing up and down.\n\n![Graph](image://a198586f-bd2c-491c-8382-823794434830 \"Parameters changing up and down.\")","13469af6-18a5-4823-ae2f-d35a29ee73c3",[1188],{"id":687,"data":1189,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":1190,"multiChoiceQuestion":1191,"multiChoiceCorrect":1193,"multiChoiceIncorrect":1195,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":1199,"matchPairsPairs":1200},[684,688,689],[1192],"Machine learning is essentially a template with parameters attached. In simple terms, what are parameters?",[1194],"Numerical values",[1196,1197,1198],"Algorithms","Strings of logic","Symbolic labels",[179],[1201],{"left":1194,"right":1202,"direction":27},"Parameters",{"id":1204,"data":1205,"type":24,"maxContentLevel":27,"version":24},"fb68e3fb-43a0-4e25-868e-36444433f942",{"type":24,"contentRole":42,"markdownContent":1206,"audioMediaId":1207},"Here's that example we promised. Let's imagine you're building an AI model which can predict the price of a pizza based on its size.\n\nYour template has parameters for all the different sizes: 10-inch pizza, 11-inch pizza, 12-inch pizza, and so on. These parameters all have prices attached, which you've estimated yourself.\n\n![Graph](image://9300acc9-f0fc-4285-aa7c-32325ceb510f \"Estimated pizza prices.\")\n\nYou then give the AI some menus from local pizza restaurants. As each menu comes through, it tweaks its parameters, and the prices of each pizza gradually go up or down.\n\nFor example, if one menu showed a 10-inch pizza for $8.50, the AI would tweak that original parameter down a couple of notches. If the next menu showed a 10-inch pizza for $9.50, it tweaks the number up again. Every time it does this, it gets closer and closer to an average price for a 10-inch pizza in your area.\n\n![Graph](image://9e5222e3-c4b8-416a-b8b9-24657cb27570 \"Adjusting prices up and down.\")","93622f33-5ccf-43ca-868c-ddf003c3a325",{"id":1209,"data":1210,"type":24,"maxContentLevel":27,"version":24,"reviews":1213},"251e4dea-38a8-46b1-9211-e6c5e7314fef",{"type":24,"contentRole":42,"markdownContent":1211,"audioMediaId":1212},"By the end, the AI will have a new set of numbers: it dropped from $10.00 to $9.00 for a 10-inch, from $11.00 to $10.00 for an 11-inch, and from $12.00 to $11.00 for a 12-inch.\n\n![Graph](image://e5bfef35-995a-49ee-a5cb-f098c43a3d8d \"Pizza prices.\")\n\nIn other words, this AI has 'learned' some average prices for pizzas being sold in your area. And if you asked it to estimate the cost of, say, a 14-inch pizza, it would hopefully give you a decent answer. In this case, maybe $13.00.\n\nYou can check if that number is accurate by looking in another pizza menu. You glance through the options, and find that this particular pizza restaurant is selling 14-inch pizzas for $13.25. That's not *exactly* the same as your AI's estimate, but all things considered, it's pretty close.","9a0cee73-47a3-454e-a06a-ea543d9932a9",[1214],{"id":1215,"data":1216,"type":67,"version":24,"maxContentLevel":27},"17b68b2f-7243-4c22-97bf-a479489f3406",{"type":67,"reviewType":37,"spacingBehaviour":24,"clozeQuestion":1217,"clozeWords":1219},[1218],"When a machine is learning, its parameters are just changing up and down.",[1220],"parameters",{"id":1222,"data":1223,"type":24,"maxContentLevel":27,"version":24,"reviews":1226},"720ea9d7-67fe-41d6-a584-8c3a9b897d77",{"type":24,"contentRole":42,"markdownContent":1224,"audioMediaId":1225},"There's actually a name for this 'gap' between the AI's prediction ($13.00) and the real-life data ($13.25). Scientists call it the **loss function**, and it's an important part of machine learning.\n\nThe smaller the loss function, the better your machine has learned. In an ideal scenario, the loss function would be non-existent – your AI would have predicted $13.25 in the first place.\n\nTo reduce the loss function for this particular AI, you could give it more menus to learn from. In theory, with every piece of data it encounters, it will get closer and closer to the 'truth'. It's not always that simple, but as a general rule, larger datasets usually lead to more effective AI models.\n\n![Graph](image://9a205d1e-fff1-4b1c-93e4-71d7062907ab \"Pizza by Valerio Capello, English Wikipedia (CC BY-SA 3.0) \u003Chttp://creativecommons.org/licenses/by-sa/3.0/>, via Wikimedia Commons\")","143bde96-0315-4b96-9c2c-2707d389c731",[1227,1242,1251],{"id":1141,"data":1228,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":1229,"multiChoiceQuestion":1230,"multiChoiceCorrect":1232,"multiChoiceIncorrect":1234,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":1238,"matchPairsPairs":1239},[1142,1143,1138],[1231],"What do we call the gap between an AI's predictions, and real-life data?",[1233],"Loss function",[1235,1236,1237],"Error margin","Deviation factor","Discrepancy measure",[179],[1240],{"left":1233,"right":1241,"direction":27},"Gap between output and real-life data",{"id":1243,"data":1244,"type":67,"version":24,"maxContentLevel":27},"44a8ec4c-5410-4862-a820-bcb5e75e4162",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":1245,"binaryCorrect":1247,"binaryIncorrect":1249},[1246],"Which of these is a general rule for machine learning?",[1248],"Larger datasets lead to smaller loss functions",[1250],"Larger datasets lead to larger loss functions",{"id":1252,"data":1253,"type":67,"version":24,"maxContentLevel":27},"2749d721-ea48-4906-8679-d3e21e48f3ed",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":1254,"activeRecallAnswers":1256},[1255],"An AI predicts that the sun will rise at 06:43. The sun actually rises at 06:47. What is the loss function?",[1257],"4 minutes",{"id":1259,"data":1260,"type":42,"version":24,"maxContentLevel":27,"summaryPage":1262,"introPage":1269,"pages":1275},"db74e4ac-2ec6-41ba-a8b1-d01d3b016868",{"type":42,"title":1261},"Supervised learning",{"id":1263,"data":1264,"type":27,"maxContentLevel":27,"version":24},"f1427bab-de36-4f39-a27a-a61975472270",{"type":27,"summary":1265},[1266,1267,1268],"Supervised learning is when you train your AI using a labeled set of data","Labeled data is essentially just a list of input-output pairs","The AI 'learns' by identifying rules and relationships between these input-output pairs",{"id":1270,"data":1271,"type":29,"maxContentLevel":27,"version":24},"2d972447-8d79-4d14-952c-5961d08daa4e",{"type":29,"intro":1272},[1273,1274],"What is the meaning of supervised learning?","When should this approach be used?",[1276,1303,1308,1331],{"id":1277,"data":1278,"type":24,"maxContentLevel":27,"version":24,"reviews":1281},"d747844d-c0a3-4c5f-85af-70b01172e141",{"type":24,"contentRole":42,"markdownContent":1279,"audioMediaId":1280},"When you want an AI to learn from a dataset, like in that pizza example we talked about last time, there are a few different approaches you can take.\n\nOne of these approaches is **supervised learning**. This is when you train your AI using a labeled set of data. When we say 'labeled', we mean that the data has all been carefully arranged into something called input-output pairs.\n\nThat's what we did with those pizzas last time. The input side of the pair was the size of each pizza (e.g. 10 inch) and the output side of the pair was the price of each pizza (e.g. $9.00).\n\nIn this example, our labeled data is essentially just a list of these size-price (input-output) pairs.","28f65da0-7464-4740-9cf7-3a9fb31b3914",[1282,1296],{"id":492,"data":1283,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":1284,"multiChoiceQuestion":1285,"multiChoiceCorrect":1287,"multiChoiceIncorrect":1289,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":1293,"matchPairsPairs":1294},[493,489,494],[1286],"What is the definition of supervised learning?",[1288],"Training AI using labeled data",[1290,1291,1292],"Training AI using unstructured data","Training AI using punishments and rewards","Manually programming an AI",[179],[1295],{"left":1288,"right":1261,"direction":27},{"id":1297,"data":1298,"type":67,"version":24,"maxContentLevel":27},"3da65ec5-757a-42d8-9733-69a99e0fb260",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":1299,"activeRecallAnswers":1301},[1300],"In supervised learning, the data needs to be manually arranged into what?",[1302],"Input-output pairs",{"id":1304,"data":1305,"type":24,"maxContentLevel":27,"version":24},"bd9b040d-d22d-4727-99b7-447bfe16bcc6",{"type":24,"contentRole":42,"markdownContent":1306,"audioMediaId":1307},"An input-output pair will hopefully have some kind of relationship or rule. In our pizza example, the rule went like this: the output number (price) was always *one less* than the input number (size).\n\n![Graph](image://974fb9ca-d2dd-4247-aaee-ca0d4c526f2c \"Pizza prices.\")\n\nIn a lot of cases, we won't actually know that relationship or rule ourselves. Instead, we want the AI to look at the input-output pairs, and search for patterns that help it to establish the relationship or rule by itself.\n\nOnce it's established the rule, it can start using it. For example, we could give an input number to our pizza AI: \"5-inch\". Using the rule above, the AI could predict the output number: \"$4.00\".\n\nIn plainer language, we could ask the AI, \"How much do you think a 5-inch pizza would cost?\" and it could reply to you, \"It will probably cost $4.00\".","6853f56a-ae06-46dd-8c63-e6628433c3fa",{"id":1309,"data":1310,"type":24,"maxContentLevel":27,"version":24,"reviews":1313},"fbc759b5-3588-4de1-ba9e-0c4e9c64df77",{"type":24,"contentRole":42,"markdownContent":1311,"audioMediaId":1312},"Here's another example. We could give our AI a bunch of emails to look at. Half of these emails are labeled 'spam', and the other half are labeled 'not spam'. This time, the input is the email itself, and the output is either 'spam' or 'not spam'.\n\nUsing this labeled data, the AI can now learn some rules or patterns that distinguish spam from non-spam. For example, it might notice that an input email with more spelling mistakes, or strange punctuation, is typically linked to an output label saying 'spam'.\n\nLater, whenever we give this AI a random email, it can use this rule to sort it. Lots of spelling mistakes? Or unusual punctuation? The email is sent to spam.\n\n![Graph](image://4cabceb9-902b-487f-adfd-c2aac2e32121 \"Spam or not spam? Image: gmcgcc, CC BY-SA 4.0 \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")","b627c77b-04cd-4b02-a3ea-658d8dc39c47",[1314],{"id":1315,"data":1316,"type":67,"version":24,"maxContentLevel":27},"364a4be5-5793-4a6e-8346-87b0c79dc841",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":1317,"matchPairsPairs":1319,"matchPairsShowExamples":6},[1318],"A supervised AI establishes a rule. The height of a dog (input) is double the length of its tail (output). Using this rule, match the pairs below:",[1320,1323,1326,1329],{"left":1321,"right":1322,"direction":27},"This dog is 50 cm tall","The dog's tail is 25 cm long",{"left":1324,"right":1325,"direction":27},"This dog is 80 cm tall","The dog's tail is 40 cm long",{"left":1327,"right":1328,"direction":27},"This dog is 20 cm tall","The dog's tail is 10 cm long",{"left":292,"right":1330,"direction":27},"The dog's tail is 100 cm long",{"id":1332,"data":1333,"type":24,"maxContentLevel":27,"version":24,"reviews":1336},"952f96f8-7681-4bdd-826f-c1360522060f",{"type":24,"contentRole":42,"markdownContent":1334,"audioMediaId":1335},"Supervised learning is great in scenarios like that email example – when there's a clear pair of inputs and outputs.\n\nHere are some other examples: when you give your AI some text in one language (input) then get a translated version (output). When you play a piece of music to your AI (input) then get the title of that song (output).\n\nAlso, that cat and dog example we mentioned earlier on. You can give it a photo of an animal (input) then get the name of that animal (output).\n\n![Graph](image://af27fc29-bc2d-4547-96eb-053ed754f436 \"Cats and dogs.\")\n\nThis isn't the only approach to machine learning. We'll look at some other types later. But it's a very common approach right now, with thousands of different uses.","12c316c2-e8e3-4121-960d-2d57fac1c02e",[1337],{"id":1338,"data":1339,"type":67,"version":24,"maxContentLevel":27},"763fb240-ffe9-4675-b248-a309c3869276",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":1340,"matchPairsPairs":1342,"matchPairsShowExamples":6},[1341],"Supervised learning works best for data with clear pairs of input and outputs. Match the pairs below:",[1343,1346,1349,1352],{"left":1344,"right":1345,"direction":27},"Movie budget (input)","Star rating (output)",{"left":1347,"right":1348,"direction":27},"Medical symptoms (input)","Disease diagnosis (output)",{"left":1350,"right":1351,"direction":27},"Size of house (input)","Market price (output)",{"left":1353,"right":1354,"direction":27},"Photo of animal (input)","Species identification (output)",{"id":1356,"data":1357,"type":42,"version":42,"maxContentLevel":27,"summaryPage":1359,"introPage":1366,"pages":1371},"ffc93e38-1ccc-45ba-9345-e5171e03b3ac",{"type":42,"title":1358},"Unsupervised learning",{"id":1360,"data":1361,"type":27,"maxContentLevel":27,"version":24},"9469463b-a890-4408-879d-e2d6bc9c97e9",{"type":27,"summary":1362},[1363,1364,1365],"Unsupervised learning involves giving AI unlabelled data to find patterns","Clustering is a technique in unsupervised learning, which sorts data into similar groups","Association is another technique which searches for relationships between data ",{"id":1367,"data":1368,"type":29,"maxContentLevel":27,"version":24},"0180b0a0-e72f-4348-8ce3-7d2f9d7b70f1",{"type":29,"intro":1369},[1370,1274],"What is the meaning of unsupervised learning?",[1372,1396,1401,1423,1428],{"id":1373,"data":1374,"type":24,"maxContentLevel":27,"version":24,"reviews":1377},"ef5c19ab-dba0-46b8-8508-c39d9f89fb66",{"type":24,"contentRole":42,"markdownContent":1375,"audioMediaId":1376},"The main alternative to supervised learning is (can you guess it?) **unsupervised learning**.\n\nThis time, you won't be giving your AI a set of nicely labeled pairs. Instead, you'll be giving it an unstructured pile of raw, unlabelled data.\n\nThis data might actually follow some interesting rules and patterns. But as a human, you don't know what they are. The data is too messy, too large, too confusing. There's certainly nothing as intuitive here as a simple input-output pair.\n\nSo you ask your AI \"are there any patterns in here?\" and see if it can learn anything useful.","c1dcb7d1-cbef-4ea3-8877-78c1c7ab6c75",[1378,1389],{"id":493,"data":1379,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":1380,"multiChoiceQuestion":1381,"multiChoiceCorrect":1383,"multiChoiceIncorrect":1384,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":1386,"matchPairsPairs":1387},[492,489,494],[1382],"What is the definition of unsupervised learning?",[1290],[1385,1291,1292],"Training AI using labelled data",[179],[1388],{"left":1290,"right":1358,"direction":27},{"id":1390,"data":1391,"type":67,"version":24,"maxContentLevel":27},"0d277827-3c7a-43ea-9c95-956576de4602",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":1392,"activeRecallAnswers":1394},[1393],"What is the main purpose of unsupervised learning?",[1395],"Identifying patterns in messy datasets",{"id":1397,"data":1398,"type":24,"maxContentLevel":27,"version":24},"9d2b8a4a-6eeb-425e-ba99-e2ee963ebc28",{"type":24,"contentRole":42,"markdownContent":1399,"audioMediaId":1400},"Typically, an unsupervised learning model will use a technique called **clustering**. This technique involves sorting data into groups based on apparent similarities and differences.\n\nIt's like giving a child a handful of marbles, then asking them to sort them. They may start to sort them by size, or color, or opacity, or weight, or whatever other patterns they come up with. An AI can do the same with data, seeking patterns, then learning from them.\n\nHere's a real world example. Imagine that you run a gym. You ask an AI to look for patterns in a database of all your members.\n\nIt finds some interesting clusters: people who live in the east of the city seem more interested in taking yoga classes, whereas people who live in the west of the city are more interested in taking spin classes. Why? You have no idea. But later, when you open a new gym in the east, you make sure it specializes in yoga.\n\n![Graph](image://450f2747-75d6-477c-9fbb-473fd3bd4f66 \"Clustered data. Image: BogdanShevchenko (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")","cbfdbd0f-6c04-4074-9fa9-9d7d766f948d",{"id":1402,"data":1403,"type":24,"maxContentLevel":27,"version":24,"reviews":1406},"99393225-f569-47d5-bfc2-cd86cec2076c",{"type":24,"contentRole":42,"markdownContent":1404,"audioMediaId":1405},"Another technique that a model might use is **association**. This involves looking for connections between pieces of data. Often, these connections are causal: if X happens, then Y will happen next.\n\nAs another example, imagine that you run a streaming service like Netflix. You ask an AI to look for patterns in a database of all your members. It finds an association: after watching a scary horror movie, it's common for people to calm down with their favorite rom com.\n\nThis knowledge helps you adjust your service – you make sure that rom coms are typically suggested after horrors.","3908f885-4b23-4887-afd1-42c4d8eda2d1",[1407,1416],{"id":1408,"data":1409,"type":67,"version":24,"maxContentLevel":27},"6b77d05a-deb5-4f9a-9c84-129543e6e90e",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":1410,"binaryCorrect":1412,"binaryIncorrect":1414},[1411],"Which unsupervised technique sorts data into groups based on similarities?",[1413],"Clustering",[1415],"Association",{"id":1417,"data":1418,"type":67,"version":24,"maxContentLevel":27},"b2c8d412-b089-4b30-928b-16d398aa67f7",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":1419,"binaryCorrect":1421,"binaryIncorrect":1422},[1420],"Which unsupervised technique looks for relationships between pieces of data?",[1415],[1413],{"id":1424,"data":1425,"type":24,"maxContentLevel":27,"version":42},"8053e1f2-4130-48b5-96c6-8171dec183d9",{"type":24,"contentRole":42,"markdownContent":1426,"audioMediaId":1427},"Unsupervised learning is incredibly powerful for exploratory analysis. It can identify all kinds of rules and patterns that you never would have thought of on your own. But remember: these rules will sometimes be bizarre and useless.\n\nYour AI might notice that people born in November like movies about dogs, while people born in September prefer movies which feature cats – but only if those cats are black. For whatever reason, this pattern might genuinely be present in your data.\n\n![Graph](image://c370f4f2-a295-4f26-87e7-aece84df2053 \"Black cat by Dmitry Makeev (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")\n\nBut is it useful? That's for you to decide.","e0818c8e-4a63-440e-8b17-d72ff0f35847",{"id":1429,"data":1430,"type":24,"maxContentLevel":27,"version":24,"reviews":1433},"5430bc56-3176-49fa-9813-0548c5c8a544",{"type":24,"contentRole":42,"markdownContent":1431,"audioMediaId":1432},"It's worth pointing out that unsupervised learning only works on massive datasets. And it's more computationally complex than supervised learning – it takes up a lot more power.\n\nThere's a trade off here. Unsupervised learning is more work for a computer, but it's much less work for a human. Supervised learning, on the other hand, is easier for the computer, but it takes time and effort on the human side to sort and label the data.\n\nA scientist will generally choose between them based on the ultimate goal of their project. If they're looking to find some unexpected patterns in a giant dataset, they'll use unsupervised. If they just want to find some simple relationships between inputs and outputs, they'll use supervised learning instead.","9ce62ffd-a232-497c-9b73-d08d012804d0",[1434],{"id":1435,"data":1436,"type":67,"version":24,"maxContentLevel":27},"ccab711e-c983-4d28-9281-c40dc037c6bd",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":1437,"multiChoiceCorrect":1439,"multiChoiceIncorrect":1443,"multiChoiceMultiSelect":21,"multiChoiceRevealAnswerOption":6},[1438],"What are some of the main drawbacks of unsupervised learning?",[1440,1441,1442],"Only works on massive datasets","Takes a lot of computing power","Sometimes comes up with strange patterns",[1444,1445],"Only works on small datasets","Takes a lot of manual setup",{"id":1447,"data":1448,"type":42,"version":24,"maxContentLevel":27,"summaryPage":1449,"introPage":1456,"pages":1461},"1f594859-93e2-4a4d-9efa-4183ac18ae31",{"type":42,"title":419},{"id":1450,"data":1451,"type":27,"maxContentLevel":27,"version":24},"4c4e2898-d816-4425-a635-e50e9209422a",{"type":27,"summary":1452},[1453,1454,1455],"Reinforcement learning uses rewards and punishments to train an AI","These rewards and punishments are actually just positive and negative values","Reinforcement learning is used for dynamic, real world situations that can't easily be summed up in a database",{"id":1457,"data":1458,"type":29,"maxContentLevel":27,"version":24},"fbd3eca3-da91-42fb-ad97-d910fe152dd3",{"type":29,"intro":1459},[1460,1274],"What is the meaning of reinforcement learning?",[1462,1479,1492,1507],{"id":1463,"data":1464,"type":24,"maxContentLevel":27,"version":24,"reviews":1467},"b8beb943-9690-45eb-a225-5c321bd3b6e9",{"type":24,"contentRole":42,"markdownContent":1465,"audioMediaId":1466},"Along with supervised learning and unsupervised learning, another popular approach to machine learning is something called **reinforcement learning**.\n\nThis one is pretty different from the others, as it doesn't require a dataset. Instead, you're going to be putting your AI in a closed environment, where it can learn via rewards and punishments.\n\nThese rewards and punishments are just positive values and negative values which are assigned to different actions. Whenever the AI performs an action, the corresponding value will tell the AI whether to perform that action again.\n\nFundamentally, it's just like operant conditioning. But instead of using rewards and punishments to train rats and monkeys, it's being used to train an AI.","34fc387d-6b08-4c10-a2dd-9a7b2adb10ba",[1468],{"id":494,"data":1469,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":1470,"multiChoiceQuestion":1471,"multiChoiceCorrect":1473,"multiChoiceIncorrect":1475,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":1476,"matchPairsPairs":1477},[492,493,489],[1472],"What is the definition of reinforcement learning?",[1474],"Training AI using rewards and punishments",[1385,1290,1292],[179],[1478],{"left":1474,"right":419,"direction":27},{"id":1480,"data":1481,"type":24,"maxContentLevel":27,"version":24,"reviews":1484},"28f721fb-e62d-4c0c-8795-19b615695fff",{"type":24,"contentRole":42,"markdownContent":1482,"audioMediaId":1483},"Reinforcement learning would be a great approach if you wanted to teach an AI model how to play a game of chess.\n\n![Graph](image://37632363-4caa-4564-9930-115482039eab \"Chess by User:Cburnett (CC BY-SA 3.0) \u003Chttp://creativecommons.org/licenses/by-sa/3.0/>, via Wikimedia Commons\")\n\nYou'd need to build an environment where the AI can play chess matches over and over and over. Each win is associated with a positive value; each loss is associated with a negative value.\n\nBy the time it's played thousands of matches, this AI might have learned which sets of moves are most likely to lead to a positive, winning outcome. It might also have learned which moves to avoid, or which tactics are too risky to pull off.\n\nReinforcement learning can also be used in other scenarios, like teaching a robot how to walk. Assign a positive value to a successful step, and a negative value if the robot falls to the floor.","bdb2f374-c7a6-46d9-b803-127b30c4642c",[1485],{"id":1486,"data":1487,"type":67,"version":24,"maxContentLevel":27},"20e1af39-480d-4f38-9aef-64ad5836a0c9",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":1488,"activeRecallAnswers":1490},[1489],"In the context of reinforcement learning, rewards and punishments are really what?",[1491],"Positive values and negative values which are assigned to different actions",{"id":1493,"data":1494,"type":24,"maxContentLevel":27,"version":24,"reviews":1497},"f1c7424b-c44b-4e80-9fd3-ce897028b392",{"type":24,"contentRole":42,"markdownContent":1495,"audioMediaId":1496},"In general, you'd use reinforcement learning if you wanted your AI to learn how to deal with dynamic, real world situations. The kinds of situations that can't easily be summed up in a database.\n\nAnother example would be training a self-driving car. Instead of getting that AI to look for patterns in data about collisions and traffic codes, it might work better to build an artificial environment where the AI can simulate driving around, and learn through trial and error.\n\n![Graph](image://27294a6d-78b9-4ed8-aae7-9cee59c6c4b3 \"Self-driving car by Grendelkhan (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")\n\nHaving said that, there might also be value in getting an AI model to learn some basic principles using supervised learning, then to graduate it to reinforcement learning later. This approach was actually how AlphaGo was trained to play the game of Go.","8b5159bf-47b1-47de-bd5f-55078ddfe1bf",[1498],{"id":1499,"data":1500,"type":67,"version":24,"maxContentLevel":27},"67d1d681-1bfc-4f9e-918e-00db4f37d037",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":1501,"binaryCorrect":1503,"binaryIncorrect":1505},[1502],"What kind of situations is reinforcement learning best suited for?",[1504],"Real world situations which can't be summed up with data",[1506],"Real world situations with lots of relevant data",{"id":1508,"data":1509,"type":24,"maxContentLevel":27,"version":24,"reviews":1512},"3f813f3f-3f88-4d22-a6bb-2706480c773c",{"type":24,"contentRole":42,"markdownContent":1510,"audioMediaId":1511},"AlphaGo initially learned from a dataset: it observed thousands of recordings of real-life moves by expert human Go players. This was a labeled dataset – the model could see when a set of moves (input) resulted in a victory (output).\n\nAfter absorbing enough basic rules and patterns through this supervised stage, the model moved to an environment where it played thousands of Go matches against versions of itself, getting better and better through reinforcement.\n\nThis combined approach worked wonders. By the end, AlphaGo had learned enough to successfully defeat the greatest Go player in the world.","91fa6f55-58ea-42e6-8d95-0cc5c1bc40a7",[1513,1525],{"id":1514,"data":1515,"type":67,"version":24,"maxContentLevel":27},"523cf760-ad5f-4e96-8126-cc7fe4639d74",{"type":67,"reviewType":15,"spacingBehaviour":24,"orderAxisType":177,"orderQuestion":1516,"orderItems":1518},[1517],"Put these stages of AlphaGo's development in order.",[1519,1521,1523],{"label":1520,"sortOrder":4},"Supervised learning (observing Go matches)",{"label":1522,"sortOrder":24},"Reinforcement learning (playing Go against itself)",{"label":1524,"sortOrder":42},"Real world play (defeating Lee Sedol)",{"id":1526,"data":1527,"type":67,"version":24,"maxContentLevel":27},"c95c116f-0078-4973-aea2-b58f3efc7e83",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":1528,"matchPairsPairs":1530,"matchPairsShowExamples":6},[1529],"In the context of reinforcement learning, match the pairs below:",[1531,1534,1537,1540],{"left":1532,"right":1533,"direction":27},"Positive value for a self-driving car","Arrive at correct destination",{"left":1535,"right":1536,"direction":27},"Negative value for a self-driving car","Arrive at wrong destination",{"left":1538,"right":1539,"direction":27},"Positive value for a voice assistant","Correctly obey command",{"left":1541,"right":1542,"direction":27},"Negative value for a voice assistant","Wrongly obey command",{"id":1544,"data":1545,"type":42,"version":24,"maxContentLevel":27,"summaryPage":1547,"introPage":1554,"pages":1560},"f78bf9aa-b641-4651-83aa-6c37ef9d0fc0",{"type":42,"title":1546},"Effectiveness of data",{"id":1548,"data":1549,"type":27,"maxContentLevel":27,"version":24},"d5a67cad-1f6a-48e8-ad73-9a758aa5dedf",{"type":27,"summary":1550},[1551,1552,1553],"Availability of data is crucial for machine learning models to be effective","Ideally, you'll need at least ten pieces of data for every parameter in your model","Lack of data can hinder AI development, especially when you're trying to use machine learning",{"id":1555,"data":1556,"type":29,"maxContentLevel":27,"version":24},"7e0c4ba5-2426-42fe-a779-b8731649bc9b",{"type":29,"intro":1557},[1558,1559],"Why is data such an important part of machine learning?","Is it possible to develop models when you don't have data?",[1561,1577,1590,1612],{"id":1562,"data":1563,"type":24,"maxContentLevel":27,"version":24,"reviews":1566},"a1975951-7b87-4dca-870d-8713d7e8b652",{"type":24,"contentRole":42,"markdownContent":1564,"audioMediaId":1565},"A lot of people are (rightly) excited about modern machine learning.\n\nAs we talked about earlier, it lets us build complex, powerful models without needing to programme all the rules of those models by hand. It also lets us build flexible, adaptive models, which can change and evolve, and handle unexpected problems.\n\nBut there's one important challenge that we need to be aware of: **availability of data**.\n\n![Graph](image://e3dcb218-1053-4f01-9fd3-7f22c128942a \"Data. By Chiffre01 (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")\n\nWhile reinforcement learning doesn't need a dataset, supervised learning and unsupervised learning do. And this can't just be any dataset. It needs to be relevant to the task that you want your AI to be learning.","8a507a33-67aa-4f9f-8f9e-11ac4c59d70f",[1567],{"id":1568,"data":1569,"type":67,"version":24,"maxContentLevel":27},"8c8f2778-035b-4c98-8bd9-b8046c752e2c",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":1570,"multiChoiceCorrect":1572,"multiChoiceIncorrect":1573,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[1571],"Which of these is a major challenge for supervised and unsupervised learning?",[173],[1574,1575,1576],"Finding large enough datasets","Finding quality datasets","Finding relevant datasets",{"id":1578,"data":1579,"type":24,"maxContentLevel":27,"version":24,"reviews":1582},"5b12189e-01b1-49bf-8f17-e2f544651161",{"type":24,"contentRole":42,"markdownContent":1580,"audioMediaId":1581},"As well as a relevant dataset, you'll also need a *large* dataset. AI learns best when it's able to repeat things thousands and thousands of times.\n\nIt's hard to put an exact number of the amount of data that you'll need. One general rule is that you'll need ten pieces of data (for example, ten photos) for each parameter in your model. If you had a model with 200 parameters? You'd need 2000 pieces of data.\n\nThat rule is pretty arbitrary. It all depends on the model, and what you want it to learn. But these approximate numbers still help to highlight how much data is often needed.","81d55c47-f16a-4bf0-96fc-2491b8450c0e",[1583],{"id":1584,"data":1585,"type":67,"version":24,"maxContentLevel":27},"d2282dba-3d94-4621-b5bf-165d1feecd58",{"type":67,"reviewType":37,"spacingBehaviour":24,"clozeQuestion":1586,"clozeWords":1588},[1587],"As a general (but arbitrary) rule, you need ten pieces of data for every one parameter in your model.",[1589],"ten",{"id":1591,"data":1592,"type":24,"maxContentLevel":27,"version":24,"reviews":1595},"7d1272a1-4597-4127-958c-1dd874972511",{"type":24,"contentRole":42,"markdownContent":1593,"audioMediaId":1594},"If you do manage to find enough usable data, you'll want to divide it into parts. The first part will be used as **training data**, which your AI model can learn from.\n\nThe second part will be some **test data**, which you can give to the AI after the learning stage, and check how well it performs.\n\nAs another general rule: of your total data, 80% should be used for training, and 20% for testing. So if you had 2000 photos in total, you'd use 1600 to train your AI, and 400 to test it later on.","4f1a8efc-64cf-4d60-9225-9649337c40cb",[1596,1605],{"id":1597,"data":1598,"type":67,"version":24,"maxContentLevel":27},"bf7a8bdd-2dc7-4680-af4b-f3185d338f1f",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":1599,"binaryCorrect":1601,"binaryIncorrect":1603},[1600],"When training an AI, what percentage of your data should you use as training data?",[1602],"80%",[1604],"20%",{"id":1606,"data":1607,"type":67,"version":24,"maxContentLevel":27},"e17540b4-6039-4c7a-ad24-cb16a1a4eb7e",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":1608,"binaryCorrect":1610,"binaryIncorrect":1611},[1609],"When training an AI, what percentage of your data should you use as testing data?",[1604],[1602],{"id":1613,"data":1614,"type":24,"maxContentLevel":27,"version":24,"reviews":1617},"a01130c0-2aaf-4609-956b-77b51bb0e216",{"type":24,"contentRole":42,"markdownContent":1615,"audioMediaId":1616},"If you can find all that data, that's amazing news. AI scientists often talk about the **unreasonable effectiveness of data**. They're basically saying, when you *do* have enough data, it will probably be more effective than you'd ever expect.\n\nBut the opposite is true as well. When you *don't* have that data, you're seriously going to struggle. It's an important drawback to modern AI. It's still pretty hard to develop models for problems that don't have much data.\n\nImagine, for example, that you wanted to train an AI model to spot the symptoms of a rare disease. But you can only find about twenty case studies of people who have had that disease. That's nowhere near enough data – machine learning wouldn't be possible.\n\n![Graph](image://d968dfc2-1a18-4a2e-9179-6cd741e841c4 \"Medical data. Image via Pexels\")","40033b4b-6457-448a-987b-4c5e37b131bf",[1618,1625],{"id":1619,"data":1620,"type":67,"version":24,"maxContentLevel":27},"3b379141-10e9-4cfc-926e-bfd129282f57",{"type":67,"reviewType":37,"spacingBehaviour":24,"clozeQuestion":1621,"clozeWords":1623},[1622],"AI scientists often talk about the unreasonable effectiveness of data.",[1624],"effectiveness",{"id":1626,"data":1627,"type":67,"version":24,"maxContentLevel":27},"882f552d-4443-4de8-a2ff-cb27e54af335",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":1628,"activeRecallAnswers":1630},[1629],"Why might it be hard to train an AI to identify rare diseases?",[1631],"There isn't much data on rare diseases",{"id":1633,"data":1634,"type":28,"maxContentLevel":27,"version":24,"orbs":1637},"480d4d34-4f3e-4fa0-a251-f18a09851856",{"type":28,"title":1635,"tagline":1636},"Neural Networks","The great leap forward",[1638,1734,1814,1896,1990],{"id":1639,"data":1640,"type":42,"version":24,"maxContentLevel":27,"summaryPage":1642,"introPage":1649,"pages":1655},"c265f697-d4d5-4225-8e42-8527920369c1",{"type":42,"title":1641},"What are neural networks?",{"id":1643,"data":1644,"type":27,"maxContentLevel":27,"version":24},"d9762c17-1cc0-4be4-971b-d5486f2613f3",{"type":27,"summary":1645},[1646,1647,1648],"Neural networks are webs of notes connected with great complexity","Early neural networks were limited in scope, as they only had one layer of nodes","Nowadays, neural networks have multiple layers, which allow them to perform powerful processes",{"id":1650,"data":1651,"type":29,"maxContentLevel":27,"version":24},"e9a98102-1aa0-4ef2-9b7c-75e5d7a6bffd",{"type":29,"intro":1652},[1653,1654],"What sparked the idea for neural networks back in the 1940s?","Why are modern neural networks more powerful than the models back then?",[1656,1676,1712],{"id":1657,"data":1658,"type":24,"maxContentLevel":27,"version":24,"reviews":1661},"4b191e9c-5ed3-411f-9965-19693ba985bc",{"type":24,"contentRole":42,"markdownContent":1659,"audioMediaId":1660},"Earlier, we touched upon **neural networks**, and mentioned how this key innovation was a major factor in the rise of modern AI. Now, it's time to look at this technology in more detail.\n\nIt's based on an idea that first cropped up in the 1940s – that's around the same time that Alan Turing was active. It was put forward by Warren McCulloch – a professor of psychiatry – and Walter Pitts – a student mathematician.\n\nTheir idea was this: neurons in the brain could basically be viewed as binary gates, just like the ones in a computer. By extension, if you built a man-made network of binary gates, connected together with great complexity, it would potentially be able to perform the same processes as a brain.","59ee65d6-4bd9-4f6c-a774-18853d3df3c2",[1662],{"id":512,"data":1663,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":1664,"multiChoiceQuestion":1665,"multiChoiceCorrect":1667,"multiChoiceIncorrect":1669,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":1672,"matchPairsPairs":1673},[510,507,511],[1666],"In the 1940s, who proposed the idea that neurons were similar to binary gates?",[1668],"McCulloch & Pitts",[1670,289,1671],"Frank Rosenblatt","Babbage & Lovelace",[179],[1674],{"left":1668,"right":1675,"direction":27},"Proposed theory of neural networks",{"id":1677,"data":1678,"type":24,"maxContentLevel":27,"version":24,"reviews":1681},"1ae0ba51-b4e8-4472-a0b5-7680988374a4",{"type":24,"contentRole":42,"markdownContent":1679,"audioMediaId":1680},"About a decade later, in 1957, an American psychologist called Frank Rosenblatt managed to put the ideas put forward by McCullock and Pitts into practice.\n\nHe constructed a network of node-like neurons, which he referred to as the **Mark I Perceptron**. Incredibly, this network used photocells to 'look' at images, and recognize objects within them.\n\n![Graph](image://0a17b588-5b3c-447d-a796-9dd1d80845aa \"Frank Rosenblatt and the Mark I Perceptron. (Public domain), via Wikimedia Commons\")\n\nThe Mark I Perceptron was only one layer thick – imagine a 2D net of nodes, as opposed to the 3D web of a real human brain. This limited the number of connections between nodes, which in turn limited the model's potential for human-like cognitive processes.\n\n![Graph](image://0646ad9b-a104-48ed-843a-d42290992929 \"Simplified diagram of a single layer network.\")\n\nBut nowadays, thanks to hundreds of innovations, we've found ways to build multilayer networks. They're still a long way away from the complex connections of a human brain. But they have enough connections to perform some pretty powerful processes.","74df243e-15cc-4eef-a9f9-a23f97872bb0",[1682,1694,1705],{"id":335,"data":1683,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":1684,"multiChoiceQuestion":1685,"multiChoiceCorrect":1687,"multiChoiceIncorrect":1688,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":1690,"matchPairsPairs":1691},[333,330,334],[1686],"Which of these is generally thought of as the world's first neural network?",[342],[339,341,1689],"AlphaGo",[179],[1692],{"left":342,"right":1693,"direction":27},"World's first neural network",{"id":510,"data":1695,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":1696,"multiChoiceQuestion":1697,"multiChoiceCorrect":1699,"multiChoiceIncorrect":1700,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":1701,"matchPairsPairs":1702},[507,511,512],[1698],"Who constructed the Mark I Perceptron?",[1670],[1668,289,1671],[179],[1703],{"left":1670,"right":1704,"direction":27},"Constructed the first neural network",{"id":1706,"data":1707,"type":67,"version":24,"maxContentLevel":27},"57b9357b-c1af-4a34-9fb3-46b46788e5d5",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":1708,"activeRecallAnswers":1710},[1709],"What feature prevented the Mark I Perceptron from performing complex processes?",[1711],"It was only one layer thick",{"id":1713,"data":1714,"type":24,"maxContentLevel":27,"version":24,"reviews":1717},"fb58016b-6e99-43dc-bb69-5df400856280",{"type":24,"contentRole":42,"markdownContent":1715,"audioMediaId":1716},"It's worth pointing out that a neural network isn't usually a physical object. These artificial neurons aren't physical nodes linked together in a physical web.\n\nInstead, it's a **computational model**: a set of digital nodes in a digital web. Just think of it like a piece of software. You can even download some neural networks, and install them on your personal computer.\n\nPhysical neural networks (PNNs) are occasionally used as well. But as you can probably imagine, they're much more fiddly to build than their digital counterparts, and harder to run at the equivalent level of complexity.","7043c54c-2551-45f1-8b71-61fb84772821",[1718,1727],{"id":1719,"data":1720,"type":67,"version":24,"maxContentLevel":27},"acfbc9a7-36a2-4433-a4c2-a79723550582",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":1721,"binaryCorrect":1723,"binaryIncorrect":1725},[1722],"Which of these would best describe a typical neural network?",[1724],"Computational model",[1726],"Physical model",{"id":1728,"data":1729,"type":67,"version":24,"maxContentLevel":27},"781a3a13-4c35-4a33-8df0-97e73ee07a44",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":1730,"activeRecallAnswers":1732},[1731],"Why are physical neural networks (PNNs) less common than digital versions?",[1733],"They are more difficult and fiddly to build",{"id":1735,"data":1736,"type":42,"version":24,"maxContentLevel":27,"summaryPage":1738,"introPage":1745,"pages":1751},"67b859fc-bef1-44ca-a76b-99f3d52504c1",{"type":42,"title":1737},"Layers",{"id":1739,"data":1740,"type":27,"maxContentLevel":27,"version":24},"60fad00b-7bc5-4f3d-93e8-93e55c99e95c",{"type":27,"summary":1741},[1742,1743,1744],"The input layer is where a neural network receives data","The hidden layers are where the network 'thinks', as data bounces through nodes","The output layer is where the AI produces a final output",{"id":1746,"data":1747,"type":29,"maxContentLevel":27,"version":24},"7e5ed9c6-d2fc-4749-a8d8-8160af69a03a",{"type":29,"intro":1748},[1749,1750],"What are the three types of layer in a neural network?","How many layers would you expect to find in a typical neural network?",[1752,1757,1770],{"id":1753,"data":1754,"type":24,"maxContentLevel":27,"version":24},"9058f82e-9a4a-4fc1-928d-3a57a2019b1a",{"type":24,"contentRole":42,"markdownContent":1755,"audioMediaId":1756},"The layers in a modern neural network are usually arranged like this. You have an **input layer**, one or more **hidden layers**, and an **output layer**.\n\n![Graph](image://3d64ee32-129a-4527-b836-00a9537e9384 \"Simplified diagram of a neural network.\")\n\nWhen you ask an AI to do something, you're interacting with the input layer. For example, you might show it a photo of an animal, and ask it \"is this a cat or a dog?\"\n\nThe input layer will send that data down into the hidden layers. As this data bounces through the web of nodes, the network is effectively 'thinking'. Assuming this model was designed to identify cats from dogs, it will try to work out what kind of animal is present in your photograph.\n\n![Graph](image://7d0a49ab-0dcb-47bd-8d67-61b17b15241f \"Cat or dog?\")\n\nEventually, the data hits the output layer. \"It's a cat,\" the AI announces.","ae4be60b-704e-4c27-b09b-2a83edfd3189",{"id":1758,"data":1759,"type":24,"maxContentLevel":27,"version":24,"reviews":1762},"b5bdfc87-d49c-43d2-a26b-899eb8c2b4f4",{"type":24,"contentRole":42,"markdownContent":1760,"audioMediaId":1761},"Interestingly, while each hidden layer might have hundreds of nodes, an output layer could have as few as two or three.\n\nFor example, in that example model we talked about, which tells the difference between cats and dogs, there are only three possible outputs: \"it's a cat\", \"it's a dog\", or \"it's neither\". All that 'thinking' in the hidden layers is just filtering to one of those options.\n\nDepending on the nature of the input, the network will take a different path through the hidden layers. If you fed it a photo of a greyhound, for example, it would 'think' about that photo in a different way than it might think about a photo of a chihuahua.\n\nBut both paths would still lead to the same output node. The AI would announce: \"It's a dog\".\n\n![Graph](image://272f283f-fcc1-4073-af53-5b0ad58aecdf \"21.png\")","9843b798-c56a-4aff-a262-a6b7950ff53c",[1763],{"id":1764,"data":1765,"type":67,"version":24,"maxContentLevel":27},"a6a1b1a5-07d5-47ab-b5fb-742fb929f1c3",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":1766,"activeRecallAnswers":1768},[1767],"Imagine a neural network which determines whether statements are true or false. How many nodes might it have in its output layer?",[1769],"Two – one for true and one for false",{"id":1771,"data":1772,"type":24,"maxContentLevel":27,"version":24,"reviews":1775},"b4073b2a-73a3-4b31-ad16-386ef06e5a4d",{"type":24,"contentRole":42,"markdownContent":1773,"audioMediaId":1774},"That cat/dog model is just a simple example. Another neural network might have hundreds of nodes in the output layer. It depends how many possible outputs the model needs to produce.\n\nIt's the same with the number of hidden layers. A simple neural network might only have one, but a more complex model might have hundreds. As a general rule, more hidden layers mean more possible paths through the web of nodes, and more powerful decision-making processes.\n\nThis principle is what brought us some of the world's most famous AI models, like AlphaGo and ChatGPT. Supposedly, the latest version of ChatGPT (GPT-4) uses a neural network with 120 hidden layers, and an enormous number of nodes.","89713acb-cd95-432b-bc33-0c81787dbac6",[1776,1787,1805],{"id":1777,"data":1778,"type":67,"version":24,"maxContentLevel":27},"1506ad91-2c03-496e-be6b-28ca95cf2272",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":1779,"multiChoiceCorrect":1781,"multiChoiceIncorrect":1783,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[1780],"Supposedly, how many hidden layers are used by ChatGPT?",[1782],"120",[1784,1785,1786],"80","40","160",{"id":1788,"data":1789,"type":67,"version":24,"maxContentLevel":27},"dd6b4791-55a7-4e46-b7f8-89774b934fd0",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":1790,"matchPairsPairs":1792,"matchPairsShowExamples":6},[1791],"What are the three main layers in a neural network?",[1793,1796,1799,1802],{"left":1794,"right":1795,"direction":27},"Input Layer","Receives some data",{"left":1797,"right":1798,"direction":27},"Hidden Layer","'Thinks' about data",{"left":1800,"right":1801,"direction":27},"Output layer","Produces an output",{"left":1803,"right":1804,"direction":27},"Logic layer","Not a real type of layer",{"id":1806,"data":1807,"type":67,"version":24,"maxContentLevel":27},"ebee58e8-cb50-4f29-982e-677325bf28ec",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":1808,"binaryCorrect":1810,"binaryIncorrect":1812},[1809],"Which of these is a general rule for neural networks?",[1811],"More hidden layers means more powerful processes",[1813],"More hidden layers means less powerful processes",{"id":1815,"data":1816,"type":42,"version":24,"maxContentLevel":27,"summaryPage":1817,"introPage":1824,"pages":1830},"223addfd-ae2f-4a3f-88e3-b9d40c3a1978",{"type":42,"title":1202},{"id":1818,"data":1819,"type":27,"maxContentLevel":27,"version":24},"f9158a92-84b9-45cd-a91a-2ff8dd79cde4",{"type":27,"summary":1820},[1821,1822,1823],"The path a neural network takes through its nodes determines the final output","Weights are attached to connections, and are used to help the model choose a path","Biases are attached to nodes, and give it an extra nudge in one direction or another",{"id":1825,"data":1826,"type":29,"maxContentLevel":27,"version":24},"f370c1d4-5a94-4aff-b7e3-86501a590e9c",{"type":29,"intro":1827},[1828,1829],"How does an AI decide which path to take through a neural network?","What do scientists mean when they talk about weights and biases?",[1831,1856,1871],{"id":1832,"data":1833,"type":24,"maxContentLevel":27,"version":24,"reviews":1836},"e426e3f9-c80e-4ea6-9a27-80c0419ceafb",{"type":24,"contentRole":42,"markdownContent":1834,"audioMediaId":1835},"So, a neural network is a series of layers. These layers are made of interconnected nodes.\n\nAnd here's an important thing to add: every connection between two different nodes has a numerical parameter attached to it. This numerical parameter is what scientists call a **weight**.\n\nAs the AI works its way through the hidden layers, following connections from node to node, these different weights will help it decide which node to jump to next. It’s more likely to choose a connection with more weight – that’s how it’s programmed to behave.\n\n![Graph](image://2b6f0b55-25e3-4948-900d-e652eae33fbe \"Simple diagram of weights.\")","49100b62-6d3b-484f-bc7b-34d16a13360a",[1837],{"id":1838,"data":1839,"type":67,"version":24,"maxContentLevel":27},"177b2812-9888-4dbd-aaa8-3b730288d77b",{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":1840,"multiChoiceQuestion":1844,"multiChoiceCorrect":1846,"multiChoiceIncorrect":1848,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":1852,"matchPairsPairs":1853},[1841,1842,1843],"a748a768-3049-4ef0-b5d3-3e7f6da73489","7d24a430-b17f-4bbd-acd5-9499fbb4731f","3bac6e62-3c67-4f61-93bb-6a8c1232651b",[1845],"In a neural network, every connection between nodes has a parameter attached. What are these parameters called?",[1847],"Weights",[1849,1850,1851],"Biases","Tokens","Nudges",[179],[1854],{"left":1847,"right":1855,"direction":27},"Parameters attached to connections",{"id":1857,"data":1858,"type":24,"maxContentLevel":27,"version":24,"reviews":1861},"eb177b1b-7e21-4627-ab51-717b62b251ae",{"type":24,"contentRole":42,"markdownContent":1859,"audioMediaId":1860},"You can think of the connections in a neural network like a tangled forest. When the network has to 'think', it's like following a path through that forest.\n\nThis path has lots of different branches. Some of them are narrow and overgrown, while others are wide and open. If you were walking, you'd probably take the open branch, just as an AI is more likely to choose a connection with more weight.\n\nThis process is essentially how a neural network makes decisions. Whichever path it takes through the web of nodes will result in a different output.","095e21d0-fa39-4406-a1c8-0ac11c06cdfa",[1862],{"id":1863,"data":1864,"type":67,"version":24,"maxContentLevel":27},"57320799-5a04-4137-afd1-dc2677160ad7",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":1865,"binaryCorrect":1867,"binaryIncorrect":1869},[1866],"If a neural network had the choice of following one of two connections, which one would you expect it to take?",[1868],"Connection with weight of 0.8",[1870],"Connection with weight of 0.2",{"id":1872,"data":1873,"type":24,"maxContentLevel":27,"version":24,"reviews":1876},"50651bcc-50dc-4b30-99c5-1d6f82ddaf98",{"type":24,"contentRole":42,"markdownContent":1874,"audioMediaId":1875},"Weights aren't the only type of parameter that you'll find in a neural network. The other main one is something called a **bias**.\n\nUnlike weights, which are attached to the connections between nodes, a bias is attached to the nodes themselves. They're basically there to give the network an extra little nudge in one direction or another.\n\n![Graph](image://eedcb2b2-45ba-4640-ba52-87394f1de5aa \"Simple diagram of weights and biases.\")\n\nSay you had two possible connections, each with a weight of 1. The network might struggle to decide which connection to follow. But the bias nudges it down the second connection. To continue with that forest analogy, it's like a little signpost: \"if in doubt, go here.\"\n\nBiases can also be negative. \"If in doubt, do *not* go here.\"","952e8c30-71c4-4973-9379-e7cc64f87e69",[1877,1888],{"id":1841,"data":1878,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":1879,"multiChoiceQuestion":1880,"multiChoiceCorrect":1882,"multiChoiceIncorrect":1883,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":1884,"matchPairsPairs":1885},[1838,1842,1843],[1881],"In a neural network, what do we call the parameters which give decision-making an extra nudge?",[1849],[1847,1850,1851],[179],[1886],{"left":1849,"right":1887,"direction":27},"Parameters attached to nodes",{"id":1889,"data":1890,"type":67,"version":24,"maxContentLevel":27},"f746d9ae-eb41-41e6-811e-7adc757970ff",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":1891,"activeRecallAnswers":1893},[1892],"In neural networks, what are weights and biases attached to?",[1894,1895],"Weights are attached to connections","Biases are attached to nodes",{"id":1897,"data":1898,"type":42,"version":24,"maxContentLevel":27,"summaryPage":1900,"introPage":1907,"pages":1913},"0fed2916-3258-47e8-bf4e-d86f9f182599",{"type":42,"title":1899},"Deep learning",{"id":1901,"data":1902,"type":27,"maxContentLevel":27,"version":24},"a3fd8b48-248b-4647-9393-56abb36eed73",{"type":27,"summary":1903},[1904,1905,1906],"Deep learning is the official name for networks that learn through many hidden layers","Neural networks use backpropagation to learn from mistakes and adjust their parameters","Deep learning models are powerful, and capable of extremely complex learning",{"id":1908,"data":1909,"type":29,"maxContentLevel":27,"version":24},"3fde3079-29a7-4db0-b3ef-494967f54a86",{"type":29,"intro":1910},[1911,1912],"Why are neural networks so good at learning?","What do scientists mean when they talk about backpropagation?",[1914,1927,1948,1963],{"id":1915,"data":1916,"type":24,"maxContentLevel":27,"version":24,"reviews":1919},"95fdb0a8-d51e-4b2d-9bd3-3ce4e554c136",{"type":24,"contentRole":42,"markdownContent":1917,"audioMediaId":1918},"Now, you might remember what we said at the start. Neural networks were the driving force behind the modern AI spring. But why are these models so important?\n\nAs it happens, these webs of nodes are *extremely* good at learning.\n\nThis learning is most effective when a neural network has lots of hidden layers. **Deep learning** is the official name for it. 'Deep' because of all those layers.\n\nRemember: when we say that a machine is 'learning', we really just mean that numerical parameters are changing. And that's exactly what happens with a neural network: the model is able to adjust its weights and biases.","41db8541-055b-417c-b146-1959bdf80462",[1920],{"id":1921,"data":1922,"type":67,"version":24,"maxContentLevel":27},"428b1aeb-2c3e-4623-8bf2-551861cab785",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":1923,"activeRecallAnswers":1925},[1924],"What do we call the type of learning that happens in a neural network with lots of layers?",[1926],"Deep Learning",{"id":1928,"data":1929,"type":24,"maxContentLevel":27,"version":24,"reviews":1932},"85cf432d-8fb4-4f6c-b29a-33937f777429",{"type":24,"contentRole":42,"markdownContent":1930,"audioMediaId":1931},"Once a neural network has performed a task, it can check the loss function afterwards. For example, if it was solving a complex math problem, how close did it get to the right answer?\n\nAfter checking the loss function, the neural network uses a technique called **backpropagation**. This is a special algorithm which travels back up the path that the AI just took through all those layers of nodes.\n\nAlong the way, it adjusts the weights and biases according to the size of the loss function. “Actually, this was a bad path to take – let's lower the weight on this one, and this one, and bump up the bias right here.\"","4608a584-2ff8-4bf9-9ba3-a56d93c08597",[1933],{"id":1143,"data":1934,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":1935,"multiChoiceQuestion":1936,"multiChoiceCorrect":1938,"multiChoiceIncorrect":1940,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":1944,"matchPairsPairs":1945},[1141,1142,1138],[1937],"What type of algorithm do neural networks use to adjust their parameters?",[1939],"Backpropagation",[1941,1942,1943],"Forward propagation","Backgradiation","Forward gradiation",[179],[1946],{"left":1939,"right":1947,"direction":27},"Algorithm which adjusts parameters",{"id":1949,"data":1950,"type":24,"maxContentLevel":27,"version":24,"reviews":1953},"6c68832f-5c01-4cfa-a477-dd2bc02bbd35",{"type":24,"contentRole":42,"markdownContent":1951,"audioMediaId":1952},"With plenty of time, and thousands of iterations, a neural network can finetune its parameters to the point that it starts reliably following the most effective path.\n\nThat's not always the *same* path. Different inputs will require different paths. That's what the model is ultimately learning – for every single input it could possibly receive, it needs to know exactly which path to take in order to produce the best and most appropriate output.\n\nImagine, for example, that you wanted your AI to tell the difference between types of fish. You input thousands of photos of different fish, and it learns the best path for each of them. If it sees a fish with *these* markings, it should take *this* path. If it sees a fish with *these* fin-shapes, it should take *this* path.\n\nEventually, it will choose the right path for every input you throw at it.\n\n![Graph](image://1515bf34-c139-4ad0-adb6-4d2d8f17eac8 \"Can you name the fish?\")","7bd7fe07-fc1f-4d22-8d44-f53d53f95342",[1954],{"id":1955,"data":1956,"type":67,"version":24,"maxContentLevel":27},"117269a9-0724-4393-a0cb-1237b53fd9a0",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":1957,"binaryCorrect":1959,"binaryIncorrect":1961},[1958],"In which direction does backpropagation move through a neural network?",[1960],"From output layer to input layer",[1962],"From input layer to output layer",{"id":1964,"data":1965,"type":24,"maxContentLevel":27,"version":24,"reviews":1968},"d2fdba64-d08c-4703-b4e9-1b0cd2387260",{"type":24,"contentRole":42,"markdownContent":1966,"audioMediaId":1967},"Just to be clear: neural networks aren't the only type of AI model which is capable of machine learning. But as things stand, they have a couple of advantages over a lot of other approaches.\n\nFirst of all, these networks are extremely versatile. You can train them to analyze data for you. You can also train them to play games, or control self-driving vehicles. You can train them to speak, or recognize images. The list goes on and on.\n\nThey're also extremely powerful, especially deep learning models. More layers mean more nodes, and more weight and biases. In other words, more detailed and complex ways for the AI to learn to behave.\n\nAccording to some numbers leaked in 2023, ChatGPT uses a neural network with more than a trillion different parameters. Just imagine how many paths you could take through such a complex neural network.","743b1634-3bb8-4d8b-a817-275320f7b99b",[1969,1976],{"id":1970,"data":1971,"type":67,"version":24,"maxContentLevel":27},"8f19c49e-1687-4c07-b6ff-f25cb1e31e2e",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":1972,"activeRecallAnswers":1974},[1973],"According to some numbers leaked in 2023, roughly how many parameters does ChatGPT's neural network have?",[1975],"More than a trillion",{"id":1977,"data":1978,"type":67,"version":24,"maxContentLevel":27},"01dd16d5-43fe-487d-b8a1-b4c711a59fc6",{"type":67,"reviewType":15,"spacingBehaviour":24,"orderAxisType":177,"orderQuestion":1979,"orderItems":1981},[1980],"Put these stages of deep learning in order:",[1982,1984,1986,1988],{"label":1983,"sortOrder":4},"The neural network performs a task",{"label":1985,"sortOrder":24},"The neural network checks the loss function",{"label":1987,"sortOrder":42},"Backpropagation travels back through the network",{"label":1989,"sortOrder":27},"The network's parameters are adjusted",{"id":1991,"data":1992,"type":42,"version":24,"maxContentLevel":27,"summaryPage":1994,"introPage":2002,"pages":2008},"aa0cd939-6a46-4208-89f0-56282efae907",{"type":42,"title":1993},"Types of neural network",{"id":1995,"data":1996,"type":27,"maxContentLevel":27,"version":24},"fb0a6beb-848f-45ed-92d0-14cbfa66e785",{"type":27,"summary":1997},[1998,1999,2000,2001],"Physical neural networks use networks of physical nodes and connections","Recurrent neural networks use a loop-back function to remember parts of a sentence","Convolutional neural networks are amazing at analyzing images","Generative adversarial networks take a pair of networks then task them to compete",{"id":2003,"data":2004,"type":29,"maxContentLevel":27,"version":24},"adf918ec-a2e6-47a4-9310-f06a1d226581",{"type":29,"intro":2005},[2006,2007],"What are some of the main types of neural networks?","When would each of these different types be used?",[2009,2024,2039,2044],{"id":2010,"data":2011,"type":24,"maxContentLevel":27,"version":24,"reviews":2014},"f4d08f0d-9523-4b70-9cdc-88598f72caa4",{"type":24,"contentRole":42,"markdownContent":2012,"audioMediaId":2013},"It's worth pointing out that there are a few different types of neural network.\n\nEarlier we mentioned **physical neural networks (PNNs)**. These are pieces of hardware, which use networks of physical nodes and connections, rather than the digital versions which are much more commonly used.\n\nAnother example is a **recurrent neural network (RNN)**. This one is actually quite simple. In a classic multilayer neural network, data is passed from hidden layer to hidden layer in one direction. Let's say from layer 1, to layer 2, to layer 3.\n\nBut in a recurrent neural network, the data will also loop back to previous layers. Effectively, this gives the network a memory – each loop reminds the previous layers what kind of data has already come through.","9657d40f-4e47-4c11-8541-f5a5f6019010",[2015],{"id":2016,"data":2017,"type":67,"version":24,"maxContentLevel":27},"323d1ac8-d4ea-4a07-a879-44f1ea6177de",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":2018,"binaryCorrect":2020,"binaryIncorrect":2022},[2019],"What is a key characteristic of a recurrent neural network (RNN)?",[2021],"Data loops back to previous layers",[2023],"Data jumps ahead to future layers",{"id":2025,"data":2026,"type":24,"maxContentLevel":27,"version":24,"reviews":2029},"b8dffd2e-d83d-4012-8d93-c91b841c523f",{"type":24,"contentRole":42,"markdownContent":2027,"audioMediaId":2028},"The loop-back function of an RNN is useful in loads of contexts.\n\nImagine, for example, that you want an AI to finish this sentence: \"The color of the sky is \\[something\\].\"\n\nIf it only remembers the final word (\"is\"), it might output something random like \"yellow\" or \"tasty\", which logically follows \"is\", but doesn't make sense in the context of the sentence as a whole.\n\nIf each word is looped back though, and 'remembered' by the network, it's more likely to give an answer that fits the context of the sentence as a whole: \"The color of the sky is blue.\"\n\n![Graph](image://ac4575bd-0ca3-4da8-a632-57a684a3de9a \"Blue sky. Image via Pexels\")","e1a905c8-bb7b-4320-84ad-2e5b354e7abf",[2030],{"id":2031,"data":2032,"type":67,"version":24,"maxContentLevel":27},"ad5fabc9-0429-4879-9998-09022b4e03fd",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":2033,"binaryCorrect":2035,"binaryIncorrect":2037},[2034],"What is the main benefit of the loop-back function used by an RNN?",[2036],"It lets the network 'remember' what kind of data has already come through",[2038],"It lets the network 'think' about data more quickly and efficiently",{"id":2040,"data":2041,"type":24,"maxContentLevel":27,"version":24},"7cde803a-78ca-491f-b129-054848e62ecc",{"type":24,"contentRole":42,"markdownContent":2042,"audioMediaId":2043},"Here's another example. This time, imagine some satellite images of a hurricane out at sea. You want your AI to calculate whether the hurricane will hit any landmasses.\n\nIf you gave the AI just one image of that hurricane, in its current position, it would be hard to predict the trajectory.\n\n![Graph](image://9742350c-46d6-48e7-92e3-44871a4cc0e1 \"One image of a hurricane. August 28 2005 NASA (Public domain), via Wikimedia Commons\")\n\nBut if you gave it ten images, showing the hurricane's progress from initial position to current position – and your AI could 'remember' these positions, in order – it would do a much better job predicting where the hurricane will go.\n\nAny time you're working with a sequence of data – be it words, or images, or something else – a recurrent approach is often more effective than a classic neural network.","af967161-0a71-4375-8bc2-3750aa533f0a",{"id":2045,"data":2046,"type":24,"maxContentLevel":27,"version":24,"reviews":2049},"fa9c1697-3924-461c-8fb5-46863bc70c34",{"type":24,"contentRole":42,"markdownContent":2047,"audioMediaId":2048},"Throughout the rest of this pathway, we'll also encounter a few other types of neural network.\n\nAnother important one is a **convolutional neural network (CNN)**. We'll take a proper look at CNNs when we get to our tile on Computer Vision, but for now, the main thing you need to know is that they're amazing at analyzing images.\n\nOne more type is an **generative adversarial network (GAN)**. We'll be learning more about this one in our tile on generative AI – but it essentially works by taking a pair of neural networks, and tasking them to compete against one another.\n\nThere are plenty of other examples. Part of the reason why neural networks are so popular is the fact they can be used in so many different ways.","023960d9-e6c7-4ceb-b121-40e889e808b1",[2050,2057],{"id":2051,"data":2052,"type":67,"version":24,"maxContentLevel":27},"84b814bc-902e-4721-8a89-4a12ec0a801c",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":2053,"binaryCorrect":2055,"binaryIncorrect":2056},[2054],"True or false: neural networks aren't particularly versatile.",[1163],[1161],{"id":2058,"data":2059,"type":67,"version":24,"maxContentLevel":27},"d10b62fd-ff80-4c07-90ca-16a621044f90",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":2060,"matchPairsPairs":2061,"matchPairsShowExamples":6},[179],[2062,2065,2068,2071],{"left":2063,"right":2064,"direction":27},"Physical neural network (PNN)","Piece of physical hardware",{"left":2066,"right":2067,"direction":27},"Recurrent neural network (RNN)","Specializes in sequential data",{"left":2069,"right":2070,"direction":27},"Convolutional neural network (CNN)","Specializes in visual data",{"left":2072,"right":2073,"direction":27},"Generative adversarial network (GAN) ","Two networks in competition",{"id":2075,"data":2076,"type":28,"maxContentLevel":36,"version":37,"orbs":2079},"08fca3f9-36ba-4c59-8f86-030db843819a",{"type":28,"title":2077,"tagline":2078},"Natural Language Processing","How do computers talk?",[2080,2176,2284,2386],{"id":2081,"data":2082,"type":42,"version":24,"maxContentLevel":27,"summaryPage":2084,"introPage":2092,"pages":2098},"c7aa91a1-8f1f-439b-bada-d0b9212e74b0",{"type":42,"title":2083},"Talking to machines",{"id":2085,"data":2086,"type":27,"maxContentLevel":27,"version":24},"f7cf5298-e0be-4043-a609-ab916e16f648",{"type":27,"summary":2087},[2088,2089,2090,2091],"NLP allows models to understand and respond to human language","NLP is a challenge for computers, because language is so unpredictable","Rule-based approaches can be used for basic NLP models, like voice assistants","Neural networks are used for more complex NLP models",{"id":2093,"data":2094,"type":29,"maxContentLevel":27,"version":24},"5f5bb2e7-ed0b-4fee-9115-19b9a632b4ef",{"type":29,"intro":2095},[2096,2097],"Why is natural language such a challenge for computers?","How do AI models overcome this challenge?",[2099,2104,2120,2135],{"id":2100,"data":2101,"type":24,"maxContentLevel":27,"version":24},"dca909d0-014d-49c2-942e-908bd104e2f8",{"type":24,"contentRole":42,"markdownContent":2102,"audioMediaId":2103},"In this pathway so far, we've traveled on a journey from the days of Charles Babbage and Ada Lovelace, through the era of symbols and expert systems, all the way to the age of modern neural networks.\n\nLater, we'll continue this journey onward, and take a look at the future of AI. But first, let's spend a little bit more time in the present. A little bit more time in the AI spring that we're literally living through right now.\n\nThis period has seen Artificial Intelligence branch out into a number of subfields. Over the next few tiles, we'll be learning all about vision, robotics, and generative AI. But first, we're going to take a look at **Natural Language Processing**.","b1efb9fe-39ab-4592-9e01-d1ae0ccd4ff0",{"id":2105,"data":2106,"type":24,"maxContentLevel":27,"version":24,"reviews":2109},"7f9671c0-9197-4714-b0dc-a43e0e585c50",{"type":24,"contentRole":42,"markdownContent":2107,"audioMediaId":2108},"Natural Language Processing (NLP) is a catch-all term which is used to describe a computer's ability to understand and respond to real, human language.\n\nThis is a pretty big deal. Imagine you wanted to build an AI which could analyze news articles, or translate novels, or identify typos in essays. It wouldn't be able to do any of those things if it didn't know how to understand and respond to language.\n\nAnd how about voice assistants? When you say: \"Alexa, play my favorite song,\" that AI needs to understand human language if it's going to know what you're asking for. It also needs to understand language if it's going to form a meaningful reply, like: \"Of course. Now playing your favorite song.\"\n\nSo yeah. NLP is important. But here's the thing: understanding language is easier said than done.","c4cb5482-e04f-439a-80d1-bcd84c577001",[2110],{"id":2111,"data":2112,"type":67,"version":24,"maxContentLevel":27},"8e55f316-cef7-4991-b253-0bae38815cd1",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2113,"multiChoiceCorrect":2115,"multiChoiceIncorrect":2116,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[2114],"In the context of AI, what does NLP stand for?",[2077],[2117,2118,2119],"Neural Language Programming","Neural Language Processing","Natural Language Programming",{"id":2121,"data":2122,"type":24,"maxContentLevel":27,"version":24,"reviews":2125},"c3ee3de6-6697-4623-a6d3-7eae8cda30c1",{"type":24,"contentRole":42,"markdownContent":2123,"audioMediaId":2124},"Every language has a set of grammar rules. But those rules are often strange and irregular. In English, you can tweak a verb like \"walk\" into a past tense version by adding -ed: \"walked\". But for another verb, \"go\", you can't add -ed. Instead, the past tense is \"went\".\n\nAnd what about slang and abbreviations? Think of the people who use \"gonna\" instead of \"going to\", or \"y'all\" instead of \"you all\".\n\nThen there's vocabulary. A language like English has more than a million words, but often the exact same word is used to mean totally different things. We call these words homonyms. For example, a good AI would need to know the difference between \"give the bell a ring\", and \"give your partner a ring\".\n\nAnd what about subtext? When we say something, there's often a secondary meaning hidden underneath words. When your friend turns up, and you say, \"That's an... interesting hat...\" what do you actually mean?\n\n![Graph](image://5d2e5000-fdd3-4637-bc58-46be3d6104ab \"Interesting hat. Image by Viktoria Borodinova (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")","f36638d5-d9ee-4bbd-b753-83fa99029013",[2126],{"id":2127,"data":2128,"type":67,"version":24,"maxContentLevel":27},"ae841cdc-4957-4d9e-b056-e8da9a736173",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":2129,"binaryCorrect":2131,"binaryIncorrect":2133},[2130],"Homonyms can confuse NLP models. But what are they?",[2132],"Words with multiple meanings",[2134],"Words with multiple spellings",{"id":2136,"data":2137,"type":24,"maxContentLevel":27,"version":24,"reviews":2140},"30b3c0da-630e-4ae2-8e03-2c00b2eb6c2a",{"type":24,"contentRole":42,"markdownContent":2138,"audioMediaId":2139},"A basic form of NLP can be achieved using rule-based approaches.\n\nImagine a simple voice assistant. It could be programmed with a list of possible commands, like \"play a song\" or \"tell me the time.\" When somebody asks it to do something, it just checks that list, then responds.\n\nAs we talked about earlier, the Eliza chatbot used a rule-based approach to simulate the dialogue of a psychoanalyst, basically just spotting key words and patterns, then generating relevant responses.\n\nBut as always, this kind of rule-based approach can only take us so far. Nowadays, the most complex NLP models are generally neural networks. With a large enough dataset of written text, they can learn all the subtleties of natural language, like nuance, slang and subtext.","a2f27f48-5fd3-4925-a4de-51efe5185f8e",[2141,2148,2158],{"id":2142,"data":2143,"type":67,"version":24,"maxContentLevel":27},"d910de87-ce97-4dbd-ac6d-06444ac6560e",{"type":67,"reviewType":37,"spacingBehaviour":24,"clozeQuestion":2144,"clozeWords":2146},[2145],"The Eliza chatbot used a rule-based approach to NLP.",[2147],"rule-based",{"id":2149,"data":2150,"type":67,"version":24,"maxContentLevel":27},"10d742dc-cf72-49e7-9d56-82d90144c0a3",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2151,"multiChoiceCorrect":2153,"multiChoiceIncorrect":2154,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[2152],"If you want your AI to learn the subtleties of natural language, which approach is most effective?",[602],[2155,2156,2157],"Rule-based","Neither are effective","Both are effective",{"id":2159,"data":2160,"type":67,"version":24,"maxContentLevel":27},"49bed229-a718-4240-b068-e0d0a5dd6647",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":2161,"matchPairsPairs":2163,"matchPairsShowExamples":6},[2162],"Why might each of these sentences be hard for an NLP model?",[2164,2167,2170,2173],{"left":2165,"right":2166,"direction":27},"\"He drank the juice\"","Irregular verb",{"left":2168,"right":2169,"direction":27},"\"They look legit\"","Slang",{"left":2171,"right":2172,"direction":27},"\"She trains on trains\"","Homonyms",{"left":2174,"right":2175,"direction":27},"\"We need to talk\"","Subtext",{"id":2177,"data":2178,"type":42,"version":37,"maxContentLevel":36,"summaryPage":2180,"introPage":2188,"pages":2194},"67db73c8-e288-4e77-b1c8-b02a124e8169",{"type":42,"title":2179},"Language pre-processing",{"id":2181,"data":2182,"type":27,"maxContentLevel":27,"version":24},"ce4d536e-1ff2-4f32-ae56-75e8c4dee1c8",{"type":27,"summary":2183},[2184,2185,2186,2187],"Language pre-processing simplifies language, and makes it easier for an AI to work with","Tokenization breaks text into clusters of letters","Stemming cuts off word endings, while lemmatization simplifies words","Stop-word removal involves deleting filler words",{"id":2189,"data":2190,"type":29,"maxContentLevel":27,"version":24},"39bfbf3e-e0d9-45d3-b68d-70c88f5550cd",{"type":29,"intro":2191},[2192,2193],"What is language pre-processing such an important part of NLP?","What are tokenization, stemming, and lemmatization?",[2195,2208,2228,2247],{"id":2196,"data":2197,"type":24,"maxContentLevel":27,"version":24,"reviews":2200},"56cdfa51-400a-4885-85f6-a3630bcae81e",{"type":24,"contentRole":42,"markdownContent":2198,"audioMediaId":2199},"To make Natural Language Processing easier, an AI will often use a technique called **language pre-processing**. In simple terms, this means cleaning the mess from a piece of natural language, and stripping it down to a simpler, more manageable form.\n\nImagine calling to Alexa: \"Hey, Alexa, can you play my favorite song again?\" Language pre-processing might cut that down to \"Alexa play favorite song\". That second version still holds all the key information from the original version, but with the messiness cleaned up, it's easier for a computer to interpret.\n\nThink of it like digging up a fossil, and brushing away all the excess soil until you're left with nothing but a nice clean bone. Or sorry, let us rephrase that: think dig up fossil, brush away soil, left with nice clean bone.","5b7105ba-eb4e-48d0-b6bf-cd9e7f8e0f54",[2201],{"id":2202,"data":2203,"type":67,"version":24,"maxContentLevel":27},"504607b3-ae2d-4479-9fc2-e55db5f04b2c",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":2204,"activeRecallAnswers":2206},[2205],"What is the purpose of language pre-processing?",[2207],"It strips language down into a simpler, more manageable form",{"id":2209,"data":2210,"type":24,"maxContentLevel":27,"version":24,"reviews":2213},"34ed80a0-cf4b-4f25-ae8b-2fa53638e9d0",{"type":24,"contentRole":42,"markdownContent":2211,"audioMediaId":2212},"**Tokenization** is a core technique in language pre-processing. It involves breaking a text into individual units. These often correspond to clusters of letters – but for the sake of clarity, let's treat a token as a single, individual word.\n\nFor example, with a sentence like \"the birds are searching for better food\", tokenization would break it down into \\[\"the\", \"birds\", \"are\", \"searching\", \"for\", \"better\", \"food\"\\].\n\n![Graph](image://68cb7cdc-59b5-4602-bfaf-8c44d8a90b4d \"Birds searching for better food. Image via Pexels\")\n\nThis step is important, because it converts a loose stream of text into tight, individual elements. Each of these elements (or tokens) can then be analyzed one-by-one. For example, the AI might associate the token \"birds\" with certain properties, like \"feathers\", \"wings\", \"beaks\".\n\nThis approach is much easier for an AI model than attempting to analyze the entire text at once.","83c6094e-de8b-4d87-8d16-fb74f06d4efd",[2214],{"id":1842,"data":2215,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":2216,"multiChoiceQuestion":2217,"multiChoiceCorrect":2219,"multiChoiceIncorrect":2220,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":2224,"matchPairsPairs":2225},[1838,1841,1843],[2218],"What do we call the clusters of letters that text can be broken into?",[1850],[2221,2222,2223],"Chunks","Segments","Strands",[179],[2226],{"left":1850,"right":2227,"direction":27},"Clusters of letters",{"id":2229,"data":2230,"type":24,"maxContentLevel":36,"version":37,"reviews":2233},"a91ba2b8-a478-4f3d-a4d4-2583123f920c",{"type":24,"contentRole":42,"markdownContent":2231,"audioMediaId":2232},"Another important pre-processing technique is **stemming**. This involves taking a word and cutting off its ending (like -ing, or -ed). This reduces it to a stem – a more basic form of the word.\n\nFor example, \\[\"the\", \"birds\", \"are\", \"searching\", \"for\", \"better\", \"food\"\\] becomes \\[\"the\", \"bird\", \"are\", \"search\", \"for\", \"better\", \"food\"\\].\n\nA related technique is **lemmatization**. Instead of removing the endings of words, it changes words into simpler versions that still mean more or less the same. It might turn \"better\" into \"good\", or convert a conjugated verb to its base grammatical form.\n\nFor example, \\[\"the\", \"birds\", \"are\", \"search\", \"for\", \"better\", \"food\"\\] becomes \\[\"the\", \"birds\", \"be\", \"search\", \"for\", \"good\", \"food\"\\].\n\nBoth of these techniques help to standardize words, which makes them easier to understand.","294f0f51-3a98-4429-bd7e-05cdba95c2d9",[2234],{"id":2235,"data":2236,"type":67,"version":37,"maxContentLevel":36},"0b752d12-941a-46db-b582-dfd337ee3b58",{"type":67,"reviewType":27,"evolvingBehavior":27,"spacingBehaviour":24,"multiChoiceQuestion":2237,"multiChoiceCorrect":2239,"multiChoiceIncorrect":2242,"multiChoiceMultiSelect":21,"multiChoiceRevealAnswerOption":6},[2238],"Which of these words have been stemmed?",[2240,2241],"Jump","Laugh",[2243,2244,2245,2246],"Pigeons","Followed","Annoying","Excellent",{"id":2248,"data":2249,"type":24,"maxContentLevel":27,"version":27,"reviews":2252},"11534725-c857-4c31-a8e0-826c90322388",{"type":24,"contentRole":42,"markdownContent":2250,"audioMediaId":2251},"One more important pre-processing technique is something called **stop-word removal**. This involves deleting any filler words (like \"and\" or \"the\") which don't add much meaning to the text.\n\nFor example, \\[\"the\", \"bird\", \"be\", \"search\", \"for\", \"good\", \"food\"\\] becomes \\[\"bird\", \"search\", \"good\", \"food\"\\].\n\nAnd there we have it. Through a series of simple pre-processing techniques, that natural sentence (\"the birds are searching for better food\") has been reduced to a format that's much simpler and easier for a computer to work with.","93d6761c-70b1-4f49-9b94-648b7d28f5ec",[2253,2270],{"id":2254,"data":2255,"type":67,"version":24,"maxContentLevel":27},"c6cd1873-6d27-4eb1-aed5-f5eb604e5560",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":2256,"matchPairsPairs":2257,"matchPairsShowExamples":6},[179],[2258,2261,2264,2267],{"left":2259,"right":2260,"direction":27},"Tokenization","Breaking text into units",{"left":2262,"right":2263,"direction":27},"Stemming","Removing endings",{"left":2265,"right":2266,"direction":27},"Lemmatization","Simplifying words",{"left":2268,"right":2269,"direction":27},"Stop-word removal","Removing filler words",{"id":2271,"data":2272,"type":67,"version":42,"maxContentLevel":27},"8c0c2948-34d4-48ac-b8ac-f0bccf8063bb",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":2273,"matchPairsPairs":2275,"matchPairsShowExamples":6},[2274],"\"The kids are paying.\" Match the pairs below:",[2276,2278,2280,2282],{"left":2259,"right":2277,"direction":27},"\"the\", \"kids\", \"are\", \"pay\", \"ing\"",{"left":2262,"right":2279,"direction":27},"\"the kid are pay\"",{"left":2265,"right":2281,"direction":27},"\"the children be pay\"",{"left":2268,"right":2283,"direction":27},"\"kids paying\"",{"id":2285,"data":2286,"type":42,"version":42,"maxContentLevel":27,"summaryPage":2288,"introPage":2295,"pages":2301},"36053661-13d6-46fd-9d65-71520adfde45",{"type":42,"title":2287},"Feature extraction",{"id":2289,"data":2290,"type":27,"maxContentLevel":27,"version":24},"4238e408-073b-49f5-851e-5f92c6c04bdb",{"type":27,"summary":2291},[2292,2293,2294],"Feature extraction can turn a piece of language into a numerical model","Feature extraction is essential for tasks that involve longer, more complex text","Word embedding plots tokens in a graph, which an AI can more easily interpret",{"id":2296,"data":2297,"type":29,"maxContentLevel":27,"version":24},"aab9e511-c7cc-4231-98cb-6d36591dd892",{"type":29,"intro":2298},[2299,2300],"Why is feature extraction such an important part of NLP?","How do AI models turn language into numerical models?",[2302,2307,2320,2339],{"id":2303,"data":2304,"type":24,"maxContentLevel":27,"version":24},"b44ffe35-41e4-4934-98bd-0531db26c43b",{"type":24,"contentRole":42,"markdownContent":2305,"audioMediaId":2306},"Last time, we saw how language pre-processing could be used to turn a piece of natural language into a cleaner, simpler form. This makes it easier for an AI to identify meaning.\n\nFor example, if someone asked a voice assistant \"What's the weather like tomorrow?\", it could pre-process that text into \\[\"weather\", \"tomorrow\"\\]. The word \"weather\" tells the model that the user wants a forecast, while \"tomorrow\" gives a clear timeframe.\n\n![Graph](image://8bb13d31-1b58-4f31-bde0-10c402029aec \"Tomorrow's weather? Image by Tomasz Sienicki, [user: tsca, mail: tomasz.sienicki at gmail.com], tsca (CC BY 3.0) \u003Chttps://creativecommons.org/licenses/by/3.0>, via Wikimedia Commons\")\n\nIt won't always be that straightforward, though.\n\nInstead of giving a simple voice command, imagine using an AI to analyze the contents of an online article. Even with the help of language pre-processing, the AI still ends up with a body of words that it doesn't really know what to do with. In cases like this, the AI will need to use an approach which scientists call **feature extraction**.","2ffb77ea-3187-45b0-9f52-4454bdd613ed",{"id":2308,"data":2309,"type":24,"maxContentLevel":27,"version":24,"reviews":2312},"f5a30c44-2c29-4b04-9dd9-b54022a06456",{"type":24,"contentRole":42,"markdownContent":2310,"audioMediaId":2311},"Feature extraction is a way for an AI to look at a piece of complicated language, like an article or a book, then pick out the most important features.\n\nThere are a few different ways to go about this. But they're all based around a similar principle. The AI needs to turn the natural language into some kind of graph, or statistical model – something that uses numbers. After all, computers work best with numbers. That's their equivalent of language.\n\nIn other words, this is a translation exercise. Feature extraction takes human language (text), and translates it into some kind of numerical model.","45e529b9-dc6b-4108-970d-5dbb0384a157",[2313],{"id":2314,"data":2315,"type":67,"version":24,"maxContentLevel":27},"352ce636-dd1c-485a-b5ad-285ca3fa5498",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":2316,"activeRecallAnswers":2318},[2317],"Feature extraction takes human language (text), and translates it into what?",[2319],"A numerical graph or model",{"id":2321,"data":2322,"type":24,"maxContentLevel":27,"version":42,"reviews":2325},"4bb485dc-2f44-4a92-a533-a767689bd959",{"type":24,"contentRole":42,"markdownContent":2323,"audioMediaId":2324},"One common approach to feature extraction is **word embedding**.\n\nThis approach starts by taking a piece of text. For example, an online article. After pre-processing the text, the AI will take all the tokens in the article, then plot them in a multidimensional graph.\n\nIn this graph, every token is represented as a different point. And the AI is able to derive patterns and meanings from their positions. For example, related tokens might be clustered together, while unrelated tokens are positioned further apart.\n\nWe're heavily over-simplifying here. Word embedding is extremely complex. But for the purpose of this pathway, that's the general idea.","bfb8e1e9-d039-4140-94b8-506b6c2f85a7",[2326],{"id":714,"data":2327,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":2328,"multiChoiceQuestion":2329,"multiChoiceCorrect":2331,"multiChoiceIncorrect":2333,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":2335,"matchPairsPairs":2336},[709,712,713],[2330],"What term describes the plotting of tokens in a multidimensional graph?",[2332],"Word embedding",[2259,2265,2334],"Knowledge graphing",[179],[2337],{"left":2332,"right":2338,"direction":27},"Tokens plotted in a graph",{"id":2340,"data":2341,"type":24,"maxContentLevel":27,"version":24,"reviews":2344},"da7ffe95-876a-4304-a35d-946a8653b6f2",{"type":24,"contentRole":42,"markdownContent":2342,"audioMediaId":2343},"Let's take a look at an example.\n\nImagine you did use word embedding on an article. Afterwards, you end up with a multidimensional graph where \"AI\" is clustered with \"exciting\" and \"revolutionary\". But for another article, you end up with a graph where \"AI\" is clustered with \"worrying\" and \"dangerous\".\n\nWhen the AI interprets both of these graphs, it should be able to identify which article is pro-AI, and which article is a lot less keen. There's actually a name for this: **sentiment analysis**. That's when a model is used to analyze the general mood of a text.\n\nThis kind of process wouldn't be possible without feature extraction. As we said, you don't really need it to interpret simple text (\"What's the weather like tomorrow?\") but for longer, more complex NLP tasks, it's a really useful approach.\n\n![Graph](image://63a79541-0fac-40dc-b000-7c12e4abaf3b \"A real example of word embedding. Generated using 19th century literature by Siobhán Grayson (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")","dec50006-82ad-4e6d-b01f-f441714f30c9",[2345,2356,2363,2377],{"id":2346,"data":2347,"type":67,"version":24,"maxContentLevel":27},"c8b2d39a-bde7-40a5-81f3-cde8912cdf55",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2348,"multiChoiceCorrect":2350,"multiChoiceIncorrect":2353,"multiChoiceMultiSelect":21,"multiChoiceRevealAnswerOption":6},[2349],"Which of these statements are generally true of word embedding?",[2351,2352],"Related tokens are clustered together","Unrelated tokens are further apart",[2354,2355],"Unrelated tokens are clustered together","Related tokens are further apart",{"id":2357,"data":2358,"type":67,"version":24,"maxContentLevel":27},"380c1556-01d5-4a33-a913-f65df7e60286",{"type":67,"reviewType":37,"spacingBehaviour":24,"clozeQuestion":2359,"clozeWords":2361},[2360],"When an AI is used to analyze the general mood of a text, we call it sentiment analysis.",[2362],"sentiment",{"id":2364,"data":2365,"type":67,"version":24,"maxContentLevel":27},"834f2518-b141-4b42-b47a-9f0beed97371",{"type":67,"reviewType":15,"spacingBehaviour":24,"orderAxisType":177,"orderQuestion":2366,"orderItems":2368},[2367],"Put these different stages of word embedding in order:",[2369,2371,2373,2375],{"label":2370,"sortOrder":4},"Text is received",{"label":2372,"sortOrder":24},"Text is pre-processed",{"label":2374,"sortOrder":42},"Tokens are plotted",{"label":2376,"sortOrder":27},"Text is analyzed",{"id":2378,"data":2379,"type":67,"version":24,"maxContentLevel":27},"2f244cff-bb57-4f08-8f00-d38cd7c5504c",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":2380,"binaryCorrect":2382,"binaryIncorrect":2384},[2381],"When is feature extraction used?",[2383],"To interpret long and complex pieces of language",[2385],"To interpret short and simple pieces of language",{"id":2387,"data":2388,"type":42,"version":42,"maxContentLevel":36,"summaryPage":2390,"introPage":2398,"pages":2404},"7cb11cd9-cf33-4ef0-93ae-23e336e5369f",{"type":42,"title":2389},"Generation",{"id":2391,"data":2392,"type":27,"maxContentLevel":27,"version":24},"960a16f2-b96d-4484-a731-eaca8d812194",{"type":27,"summary":2393},[2394,2395,2396,2397],"Natural Language Generation is the process of generating language","Rule-based templates are a basic approach, while statistical models are more complex","Statistical models predict which words are most likely to occur together","Large Language Models (LLMs) like ChatGPT use this statistical approach",{"id":2399,"data":2400,"type":29,"maxContentLevel":27,"version":24},"d4dc0943-928e-4bdf-9bb5-bbd3c4e9f9b9",{"type":29,"intro":2401},[2402,2403],"How do NLP models generate language?","Why are statistics so important to models like ChatGPT?",[2405,2433,2450,2463],{"id":2406,"data":2407,"type":24,"maxContentLevel":27,"version":24,"reviews":2410},"8cda3475-2c65-4fe3-81db-52b4503aad95",{"type":24,"contentRole":42,"markdownContent":2408,"audioMediaId":2409},"Natural Language Processing (NLP) isn't just about turning human language into a form that computers can understand. It's also about the opposite process: turning computer language into a form that humans can understand.\n\nJust think about a chatbot or a voice assistant. It might know what you mean when you ask it, \"What's the best way to boil an egg?\" but that's not very useful if it doesn't know how to reply.\n\n![Graph](image://5f6faa4f-9a1e-4ca8-ac49-a127f2ee9b74 \"Boiled eggs. Image: (CC0) \u003Chttp://creativecommons.org/publicdomain/zero/1.0/deed.en>, via Wikimedia Commons\")\n\nThis process of turning computer language into human language is called **Natural Language Generation (NLG)**. Meanwhile, all that stuff we learned about earlier – language pre-processing and feature extraction – is generally known as **Natural Language Understanding (NLU)**.","057fd7ff-5e47-455a-a03a-29e3285f16e2",[2411,2422],{"id":2412,"data":2413,"type":67,"version":24,"maxContentLevel":27},"0eaee819-6b74-423b-b815-0df66c3b4adb",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2414,"multiChoiceCorrect":2416,"multiChoiceIncorrect":2418,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[2415],"In the context of AI, what does NLG stand for?",[2417],"Natural Language Generation",[2419,2420,2421],"Neural Language Generation","Natural Language Graphing","Neural Language Graphing",{"id":2423,"data":2424,"type":67,"version":24,"maxContentLevel":27},"c3a941d4-0d3a-4f23-9d3c-8b59101261fe",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2425,"multiChoiceCorrect":2427,"multiChoiceIncorrect":2429,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[2426],"In the context of AI, what does NLU stand for?",[2428],"Natural Language Understanding",[2430,2431,2432],"Neural Language Understanding","Natural Language Usage","Neural Language Usage",{"id":2434,"data":2435,"type":24,"maxContentLevel":27,"version":24,"reviews":2438},"45ee7bd0-9036-4f49-bdfb-8bcbc3304643",{"type":24,"contentRole":42,"markdownContent":2436,"audioMediaId":2437},"Essentially, the process of NLG is the reverse of NLU. The computer will start with some numbers, data, graphs or tables, then translate this data into coherent words and sentences.\n\nThere are a few different ways to go about this, though.\n\nThe most basic approach is a **rule-based template**. This is where an AI is programmed with a set of template sentences, which function like fill-in-the-blanks. For example: \"The weather today is \\[condition\\] with a high of \\[temperature\\].\"\n\nUsing a template like this one, it's pretty easy for the AI to look at some data (like a weather report) then output a meaningful sentence. But this approach is pretty limited. Apart from those template sentences, the AI can't say anything else.","c0c84e87-4383-4e6f-a837-1ceaedb35881",[2439],{"id":2440,"data":2441,"type":67,"version":24,"maxContentLevel":27},"870acbce-15f2-4326-92ec-84d164b75412",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2442,"multiChoiceCorrect":2444,"multiChoiceIncorrect":2446,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[2443],"\"The time right now is [time].\" This is an example of what simple type of NLG?",[2445],"Rule-based template",[2447,2448,2449],"Statistical model","Narrow template","Probabilistic model",{"id":2451,"data":2452,"type":24,"maxContentLevel":27,"version":24,"reviews":2455},"da912929-87d5-4afa-b945-9cfc46b5f80b",{"type":24,"contentRole":42,"markdownContent":2453,"audioMediaId":2454},"That rule-based template is a form of symbolic programming. And as you'd probably expect, more advanced approaches to Natural Language Generation will generally use machine learning techniques instead.\n\nOne of the most popular examples of this is something called a **statistical model**. This involves training an AI to identify patterns between words. More specifically, to identify which words are most likely to occur right after one another.\n\nFor example, imagine that the AI has already generated three words: \"The dog is...\" Statistically, these words are more likely to be followed by a word like \"barking\" or \"running\", rather than a word like \"flying\" or \"delicious\".\n\n![Graph](image://11da59c4-2e32-40ea-a0e2-604eb391ea09 \"A real example of a statistical. By Idéalités (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")","9a28fc8f-b221-4c2f-b176-c2e15e1c02b0",[2456],{"id":2457,"data":2458,"type":67,"version":24,"maxContentLevel":27},"bf5ff86f-e486-4f58-a2c4-9f742eebc19e",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":2459,"activeRecallAnswers":2461},[2460],"How does a statistical NLG model work?",[2462],"It knows which words are most likely to follow one another",{"id":2464,"data":2465,"type":24,"maxContentLevel":36,"version":42,"reviews":2468},"811514c9-bcd9-4165-b276-b3ea5a1af740",{"type":24,"contentRole":42,"markdownContent":2466,"audioMediaId":2467},"Like with any machine learning, statistical models are only as good as their data. But luckily, there's a lot of natural language out there. Think of all the millions of books that you could give to an AI model, or the billions of webpages, or news articles.\n\nGoogle Translate uses a neural network which was trained to predict likely sequences of words after analyzing a dataset of texts in various languages. ChatGPT uses a neural network which was trained using texts from the internet.\n\nThere's actually a name for NLG models like ChatGPT: Large Language Models (LLMs). They're so good at generating natural language, that they can actually pass the Turing Test.","48be79b9-78e2-4175-bff5-2dfc33e73ee2",[2469,2478,2496],{"id":2470,"data":2471,"type":67,"version":24,"maxContentLevel":27},"e3789374-f567-4e3b-91e0-c19259abaeab",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":2472,"binaryCorrect":2474,"binaryIncorrect":2476},[2473],"Why is it relatively easy to train a statistical NLG model?",[2475],"There's a lot of data (books etc.) that you can train it on",[2477],"These models learn faster than other types of neural network",{"id":2479,"data":2480,"type":67,"version":42,"maxContentLevel":36},"a4c01933-bfec-4442-9774-c0db93a454d1",{"type":67,"reviewType":177,"evolvingBehavior":177,"spacingBehaviour":24,"matchPairsQuestion":2481,"matchPairsPairs":2483,"matchPairsShowExamples":6},[2482],"What kind of data might you use to train these different NLP models?",[2484,2487,2490,2493],{"left":2485,"right":2486,"direction":27},"Translation model","Books in different languages",{"left":2488,"right":2489,"direction":27},"Large language model","Articles from the internet",{"left":2491,"right":2492,"direction":27},"Chatbot","Text conversations",{"left":2494,"right":2495,"direction":27},"Voice assistant","Phone conversations",{"id":2497,"data":2498,"type":67,"version":24,"maxContentLevel":27},"a6ad8c91-3737-4146-ae35-7551360012af",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2499,"multiChoiceCorrect":2501,"multiChoiceIncorrect":2503,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[2500],"\"The cat fell off the...\" What word might a statistical model choose next?",[2502],"\"chair\" (3.4% likely)",[2504,2505,2506],"\"horse\" (0.9% likely)","\"yellow\" (0.1% likely)","\"table\" (2.5% likely)",{"id":2508,"data":2509,"type":28,"maxContentLevel":36,"version":37,"orbs":2512},"6444ca2e-c123-4189-a7d7-26bca51e8793",{"type":28,"title":2510,"tagline":2511},"Generative AI","How do computers create?",[2513,2643,2700,2796],{"id":2514,"data":2515,"type":42,"version":24,"maxContentLevel":27,"summaryPage":2517,"introPage":2525,"pages":2531},"9011ca88-b15f-44ea-b709-326543c9b0e5",{"type":42,"title":2516},"Large Language Models",{"id":2518,"data":2519,"type":27,"maxContentLevel":27,"version":24},"9bf46eb7-d453-4fab-a2eb-cd2ad207d5fd",{"type":27,"summary":2520},[2521,2522,2523,2524],"Generative AI can create something new after learning from existing data","Large Language Models, like ChatGPT, specialize in natural language generation","These models are trained on massive amounts of data, typically taken from the internet","Large Language Models use transformer architecture to improve the quality of their outputs",{"id":2526,"data":2527,"type":29,"maxContentLevel":27,"version":24},"31a74e46-74b0-4ad8-8eb8-917e8431d310",{"type":29,"intro":2528},[2529,2530],"How are Large Language Models trained to generate language?","Where do Large Language Models get their data from?",[2532,2549,2573,2598],{"id":2533,"data":2534,"type":24,"maxContentLevel":27,"version":24,"reviews":2537},"897bd1b0-60eb-41cd-b8a9-0d37386b5ea1",{"type":24,"contentRole":42,"markdownContent":2535,"audioMediaId":2536},"In the last few years, **Generative AI** (or Gen AI, for short) has become a bit of a buzzword.\n\nIt's used to describe any AI model which can study a dataset, then use what it's learned to create something totally new. For example, it might study a thousand paintings, then generate a painting of its own. Or it might study a million bedtime stories, then write a new one completely from scratch.\n\nTo give you some context, this is totally different to something like a descriptive AI, which explains and classifies data. Where descriptive AI is designed to explain, generative AI is specifically designed to **create**.","96e31dc3-f986-4ef7-a321-cd9746c012b4",[2538],{"id":2539,"data":2540,"type":67,"version":24,"maxContentLevel":27},"624bd74d-b491-4cc4-b4c4-c077d745b373",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2541,"multiChoiceCorrect":2543,"multiChoiceIncorrect":2545,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[2542],"Generative AI can study a dataset, then use what it's learned to:",[2544],"Create something new",[2546,2547,2548],"Describe the dataset","Categorize the dataset","Predict future trends",{"id":2550,"data":2551,"type":24,"maxContentLevel":27,"version":24,"reviews":2554},"0ced03ac-51bd-4d3f-b942-eaa0c957f277",{"type":24,"contentRole":42,"markdownContent":2552,"audioMediaId":2553},"The most famous example of generative AI is ChatGPT, which exploded onto the global scene towards the end of 2022. It's an example of a particular subtype of AI, which is known as a **Large Language Model (LLM)**.\n\nWe mentioned LLMs last time. They're essentially just complex neural networks which specialize in natural language. As a general rule, we call it a 'Large' Language Model when it uses a billion parameters. The latest version of ChatGPT is said to have more than a trillion.\n\nThese vast neural networks are trained on a colossal volume of data, sometimes called a corpus. This corpus might be made up of news articles, text messages, song lyrics, novels, screenplays, textbooks, emails, poems...\n\nSome of the world's most powerful LLMs were literally trained using all the data on the internet.\n\n![Graph](image://89700d4f-675f-4c9b-8239-f9ddeb49b3c7 \"Data. Image via Pexels\")","d0964f67-9cef-4454-9118-fde5c748ea83",[2555,2566],{"id":2556,"data":2557,"type":67,"version":24,"maxContentLevel":27},"d1d2dc9d-8c34-446b-b87c-6ed8e2223421",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2558,"multiChoiceCorrect":2560,"multiChoiceIncorrect":2562,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[2559],"As a general rule, a Large Language Model has how many parameters (or more)?",[2561],"One billion",[2563,2564,2565],"One thousand","One million","One trillion",{"id":2567,"data":2568,"type":67,"version":24,"maxContentLevel":27},"97f84338-fc75-4cc4-8984-6e1ee61e4edb",{"type":67,"reviewType":37,"spacingBehaviour":24,"clozeQuestion":2569,"clozeWords":2571},[2570],"LLMs are trained on a vast amount of data, which is sometimes known as a corpus.",[2572],"corpus",{"id":2574,"data":2575,"type":24,"maxContentLevel":27,"version":24,"reviews":2578},"ec0f631c-2c0d-45f8-bd71-dbe4d666ee68",{"type":24,"contentRole":42,"markdownContent":2576,"audioMediaId":2577},"You might be wondering what 'GPT' stands for: Generative Pre-trained Transformer.\n\n'Generative' and 'pre-trained' are self-explanatory. But **transformer** is a term that you might not have seen before. It refers to a specific learning framework that helps to improve the quality of an LLM's outputs.\n\nWe won't get into all the details. But in simple terms, transformer architecture uses mathematical models to keep track of the overall context of a text, rather than just the last few words it has written.\n\nFor example, if you used an LLM to write an email, and the first sentence mentioned the recipient's name, the transformer architecture helps it remember this context, and keep using that name throughout the rest of the email.\n\nIt's worth pointing out that ChatGPT isn't the only transformer LLM on the market. Anthropic's Claude is another example. So is Google's Gemini.","e572a7c0-3059-4f15-a61d-ff10b0cd96c7",[2579],{"id":2580,"data":2581,"type":67,"version":24,"maxContentLevel":27},"ba4e79ac-ecb6-4193-a7e9-dd48e8280af9",{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":2582,"multiChoiceQuestion":2586,"multiChoiceCorrect":2588,"multiChoiceIncorrect":2590,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":2594,"matchPairsPairs":2595},[2583,2584,2585],"9ee9668b-96fd-4dd0-9dcf-8f4ab8e40111","7e7ae376-1c24-4dfd-863b-9afbebc6a364","1cb45017-3c6a-430b-8344-a2c9984a91af",[2587],"What type of LLM can keep track of the overall context of a text, rather than just the last few words it has written?",[2589],"Transformer",[2591,2592,2593],"Transceiver","Performer","Perceiver",[179],[2596],{"left":2589,"right":2597,"direction":27},"LLM that keeps track of context",{"id":2599,"data":2600,"type":24,"maxContentLevel":27,"version":24,"reviews":2603},"500ffc88-e175-4d53-8750-3b448f3b5583",{"type":24,"contentRole":42,"markdownContent":2601,"audioMediaId":2602},"It has to be said: even the world's best LLMs still struggle to compete with human writing in more expressive or artistic contexts. They lack the kind of empathy and emotion which the very best writing is based on.\n\nBut for simple writing, LLMs are extremely powerful tools. You can use them to write emails, or summarize articles, or rephrase complicated text. You can use them to draft to-do lists, or write computer code, or even just have a quick chat.\n\nInterestingly, the latest version of ChatGPT has also been trained to work with images and audio. In other words, it's no longer strictly an LLM – we can call this model **multimodal**, which basically means it can perform more than one task.","f70d310a-d9fa-4c60-b119-e9799fbba8a1",[2604,2617,2627,2636],{"id":2605,"data":2606,"type":67,"version":24,"maxContentLevel":27},"4c31dd78-37ff-473f-989e-83c888ebe3d9",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2607,"multiChoiceCorrect":2609,"multiChoiceIncorrect":2613,"multiChoiceMultiSelect":21,"multiChoiceRevealAnswerOption":6},[2608],"Which of these are examples of transformer LLMs?",[2610,2611,2612],"ChatGPT (OpenAI)","Claude (Anthropic)","Gemini (Google)",[2614,2615,2616],"Stable Diffusion (Stability AI)","Midjourney (Midjourney Inc.)","DALL-E (OpenAI)",{"id":2618,"data":2619,"type":67,"version":24,"maxContentLevel":27},"f6a5c31a-483b-4352-9a58-4bc53302f4c7",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2620,"multiChoiceCorrect":2622,"multiChoiceIncorrect":2624,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[2621],"What does GPT stand for?",[2623],"Generative Pre-trained Transformer",[2625,2626,292],"Graphical Pattern Transformer","Generative Processing Tool",{"id":2628,"data":2629,"type":67,"version":24,"maxContentLevel":27},"521714f9-17eb-4d1d-b4a7-dc48d75b13a2",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":2630,"binaryCorrect":2632,"binaryIncorrect":2634},[2631],"The latest version of ChatGPT has been trained to work with images and audio, as well as text. What do we call a model like this one?",[2633],"Multimodal AI",[2635],"General AI",{"id":2637,"data":2638,"type":67,"version":24,"maxContentLevel":27},"b04c2ed8-aa7a-4de4-a442-9e6c5a7e6777",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":2639,"activeRecallAnswers":2641},[2640],"Why might an LLM struggle to compete with human writing in more expressive or artistic contexts?",[2642],"These models lack empathy and emotion",{"id":2644,"data":2645,"type":42,"version":42,"maxContentLevel":27,"summaryPage":2647,"introPage":2653,"pages":2658},"f9b2f01b-d04c-4ac1-83b4-577e035db3e8",{"type":42,"title":2646},"Case study: Turing Test",{"id":2648,"data":2649,"type":27,"maxContentLevel":27,"version":24},"09353fb5-0670-4152-a6fd-a25b27882e42",{"type":27,"summary":2650},[2651,2652],"In a recent Turing Test study, ChatGPT was misidentified as a human 54% of the time ","In another Turing Test study, to Eliza which was only misidentified 22% of the time",{"id":2654,"data":2655,"type":29,"maxContentLevel":27,"version":24},"308f6655-e382-42b6-808f-7f595a29916f",{"type":29,"intro":2656},[2657],"Can you tell the different between ChatGPT and a human?",[2659,2664,2669,2674],{"id":2660,"data":2661,"type":24,"maxContentLevel":27,"version":24},"41862024-9713-42aa-ae9f-ec082c8de46c",{"type":24,"contentRole":42,"markdownContent":2662,"audioMediaId":2663},"Before moving on from LLMs, here's an interesting question. Is an AI model like ChatGPT capable of passing the Turing Test? In this orb, we'll find out firsthand.\n\nThere are a few variations of the Turing Test, but we'll be using one that looks like this. A machine and a human have a conversation. A second human (that's you!) reads the messages. At the end, you'll be asked a question: of the two participants, can you tell which one was a machine?\n\nThe conversation will start on the page after this one. The participants are labeled Person A and Person B. But just to repeat: only one is a real person.","f5dfac3a-691d-4a30-a7d6-a924e9d75c82",{"id":2665,"data":2666,"type":24,"maxContentLevel":27,"version":24},"3b554e91-97ad-4613-a074-cf13dafef9a1",{"type":24,"contentRole":42,"markdownContent":2667,"audioMediaId":2668},"**Person A**: \"Hey, quick question – if robots had favorite movies, what kind of films do you think they’d be into?\"\n\n**Person B**: \"Interesting. Honestly, I have no idea. Science fiction, maybe? They'd probably like seeing themselves on screen.\"\n\n**Person A**: \"Yeah, I can see that. Probably big fans of The Terminator or Wall-E, right? Imagine a robot getting super invested in robot characters.\"\n\n**Person B**: \"Ha, yeah. Like, deciding that the Terminator was the ideal role model. Or developing a crush on Wall-E.\"\n\n**Person A**: \"Exactly! A robot trying to act all cool like the Terminator... but then secretly fangirling over Wall-E’s personality. That would be hilarious.\"\n\n**Conversation continued on next page.**","382aaddd-18a3-4429-8f85-232d0452d779",{"id":2670,"data":2671,"type":24,"maxContentLevel":27,"version":24},"9ae2319a-dc54-4c08-b1e2-d56fc451145e",{"type":24,"contentRole":42,"markdownContent":2672,"audioMediaId":2673},"**Person B**: \"Yep, it sure would be! How about you, though? Do you agree that robots would like science fiction films, or is there another genre they might prefer?\"\n\n**Person A**: \"Hmm, I feel like robots might also get into documentaries, you know? Something all factual and logical, like ‘How It's Made.’ Maybe they’d enjoy learning about how the world works – or how they work.\"\n\n**Person B**: \"I like this. It's like a human watching a documentary about the digestive system. It's important to know how your body works.\"\n\n**Conversation end.**\n\nThere you have it. A conversation between a machine and a human. But which one is which? Take a moment to think. Make a guess in your mind. When you're ready to find out the actual answer, please continue to the page after this one.","34df030b-f3a3-429a-a99a-b24a7a8e7aba",{"id":2675,"data":2676,"type":24,"maxContentLevel":27,"version":42,"reviews":2679},"476ba5da-53ba-4729-a2f3-fd0d58e43363",{"type":24,"contentRole":42,"markdownContent":2677,"audioMediaId":2678},"Here's the answer: Person A was actually ChatGPT, while person B was a real-life human.\n\nMaybe you guessed that correctly. Maybe you got it wrong. But either way, it probably wasn't easy. Would you have been confident enough to bet your savings on it?\n\nJust to be clear, none of this means that ChatGPT is sentient, or conscious, or an Artificial General Intelligence. It's still nothing more than an Artificial Narrow Intelligence using a neural network to mimic natural language.\n\nBut it does this very effectively. In one recent study, using a similar test, GPT was misidentified as a human 54% of the time. Compare that to something like the Eliza chatbot, way back in the 1960s. In the same test, it was only misidentified 22% of the time.","9d9aeb87-3567-4688-9aad-f62917d01512",[2680,2691],{"id":2681,"data":2682,"type":67,"version":42,"maxContentLevel":27},"21ebd094-6e73-4bc9-8210-7694be22a0e0",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2683,"multiChoiceCorrect":2685,"multiChoiceIncorrect":2687,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[2684],"In a recent Turing Test study, how often was ChatGPT misidentified as a human?",[2686],"54%",[2688,2689,2690],"22%","45%","60%",{"id":2692,"data":2693,"type":67,"version":42,"maxContentLevel":27},"badc9a04-c019-45d7-b439-87f5b72c8609",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2694,"multiChoiceCorrect":2696,"multiChoiceIncorrect":2697,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[2695],"In a 1960s Turing Test study, how often was the Eliza chatbot misidentified as a human?",[2688],[2686,2698,2699],"30%","40%",{"id":2701,"data":2702,"type":42,"version":37,"maxContentLevel":36,"summaryPage":2704,"introPage":2712,"pages":2718},"a15d06f6-c5bf-46bc-a5d3-d002cbc46fb1",{"type":42,"title":2703},"Text-to-Image models",{"id":2705,"data":2706,"type":27,"maxContentLevel":27,"version":24},"ed0d5a90-7cd4-4801-8259-027858d546b6",{"type":27,"summary":2707},[2708,2709,2710,2711],"Text-to-Image models can generate images from text prompts","Generative Adversarial Networks are a popular way to train Text-to-Image models","Instead of 'seeing' images, these models process images as patterns of numbers","They can be used to generate digital paintings, product blueprints, and more",{"id":2713,"data":2714,"type":29,"maxContentLevel":27,"version":24},"91658a78-52b4-4c67-b91a-91a142e8a177",{"type":29,"intro":2715},[2716,2717],"How are Text-to-Image models trained to generate images?","Can these models actually 'see' the images they're generating?",[2719,2732,2760,2765,2791],{"id":2720,"data":2721,"type":24,"maxContentLevel":27,"version":24,"reviews":2724},"ddbea2bd-4eb3-4a0d-9c5c-7e62d0d9bfc1",{"type":24,"contentRole":42,"markdownContent":2722,"audioMediaId":2723},"Large Language Models, like ChatGPT, are probably the most striking and well known type of modern generative AI. But not far behind is another type: **Text-to-Image models**.\n\nYou've probably heard of at least one of them. Midjourney, DALL-E, Stable Diffusion...\n\nThese models can take a piece of text (a prompt) and use it to generate a picture. For example, if you prompted it to generate \"a picture of purple mountain,\" that's exactly what it would do.\n\n![Graph](image://7d7e7735-3884-44aa-be57-43e8c3afca68 \"A picture of a purple mountain.\")\n\nJust like those Large Language Models we looked at last time, these text-to-image models are extremely powerful and complex. The neural network behind the DALL-E model uses several billion parameters.","12742a18-7f6c-4c87-a962-d2361e0d2951",[2725],{"id":2726,"data":2727,"type":67,"version":24,"maxContentLevel":27},"e15171c9-cffa-4a8d-8d07-92dbed3eba88",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2728,"multiChoiceCorrect":2730,"multiChoiceIncorrect":2731,"multiChoiceMultiSelect":21,"multiChoiceRevealAnswerOption":6},[2729],"Which of these are examples of Text-to-Image models?",[2614,2615,2616],[2610,2611,2612],{"id":2733,"data":2734,"type":24,"maxContentLevel":36,"version":37,"reviews":2737},"1580afff-9794-4c79-af6a-72fdc279b6c3",{"type":24,"contentRole":42,"markdownContent":2735,"audioMediaId":2736},"Training a Text-to-Image model will generally involve a massive dataset of images with descriptive captions.\n\nThis is a great example of **supervised learning**: training a model using a labeled set of input-output pairs. The descriptive caption is the input, while the image is the output.\n\nThis approach allows the model to learn the relationship between words (\"a ginger cat beside a book on a windowsill\") and visuals (the image below).\n\n![Graph](image://2e8357de-d3be-4c97-9072-52712b73423a \"A ginger cat beside a book on a windowsill.\")\n\nIt's worth pointing out that a dataset like this isn't easy to put together. One example is Microsoft's Common Objects in Context (COCO), which features more than 100,000 different images, each with five descriptive captions. All of these captions had to be manually written by a team of human annotators.","81493ee2-53c7-4ccb-b062-7e3866724a08",[2738,2747],{"id":2739,"data":2740,"type":67,"version":24,"maxContentLevel":27},"0330f9f3-29a9-47ed-80d7-c7765e0dd0d5",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":2741,"binaryCorrect":2743,"binaryIncorrect":2745},[2742],"Text-to-Image models are typically trained using a labelled dataset of input-output pairs. What are these pairs?",[2744],"Descriptive caption (input), visual image (output)",[2746],"Visual image (input), descriptive caption (output)",{"id":2748,"data":2749,"type":67,"version":37,"maxContentLevel":36},"46ae3603-f6ce-4c73-afe5-9f69dce492d5",{"type":67,"reviewType":27,"evolvingBehavior":27,"spacingBehaviour":24,"multiChoiceQuestion":2750,"multiChoiceCorrect":2752,"multiChoiceIncorrect":2756,"multiChoiceMultiSelect":21,"multiChoiceRevealAnswerOption":6},[2751],"Which of these statements are true of Microsoft's Common Objects in Context dataset?",[2753,2754,2755],"Contains 100,000 images","Every image has 5 captions","The captions were written manually",[2757,2758,2759],"Contains 10,000 images","Every image has 10 captions","The captions were AI generated",{"id":2761,"data":2762,"type":24,"maxContentLevel":27,"version":24},"6638c6c4-502c-4240-bce9-b0e3c5b285e0",{"type":24,"contentRole":42,"markdownContent":2763,"audioMediaId":2764},"Text-to-Image models can also be built using something called a **Generative Adversarial Network (GAN)**.\n\nThis approach involves a pair of neural networks which are designed to work against each other. The first neural network (the Generator) produces AI generated images. The second network (the Discriminator) looks at these images, along with some real-life images, and tries to guess which ones are which.\n\nIt's almost as though the two networks are playing a game. If the Generator manages to fool the Discriminator, it gets positive feedback. If it fails, the Discriminator gets positive feedback. Over time, both networks get better and better at generating and detecting AI generated images.\n\nThe GAN approach isn't only used for Text-to-Image models. It can also be used to generate text. But it seems to work especially well for images.","02f108fd-e74c-4f90-bfcd-08f2a955dcd9",{"id":2766,"data":2767,"type":24,"maxContentLevel":27,"version":24,"reviews":2770},"4834558b-cbbf-4bc3-9a2f-bfd0b982014e",{"type":24,"contentRole":42,"markdownContent":2768,"audioMediaId":2769},"Now, it's important to remember that when a Text-to-Image model learns to generate images, it's not actually 'aware' of what those images look like. At least, not in the same way as a human would be if they were drawing, or taking a photo.\n\nInstead, every image is converted into some kind of numerical structure which the computer is able to work with. There are different ways to go about this, but one simple example would be a map of pixels, each with a number that describes that pixel's color.\n\n![Graph](image://3a15386b-e22b-47bb-960c-5f8f9089ff0d \"Simplified map of pixels.\")\n\nIn other words, what's happening behind the scenes of a Text-to-Image model is similar to an LLM. An image is treated as a complex pattern of numbers, exactly the same as an LLM might treat a piece of text.","ea17fe75-0977-4e32-96bf-933b7fa990ca",[2771,2778],{"id":2772,"data":2773,"type":67,"version":24,"maxContentLevel":27},"e5909c81-a2a0-4be3-89fb-fa8fd1b250fb",{"type":67,"reviewType":37,"spacingBehaviour":24,"clozeQuestion":2774,"clozeWords":2776},[2775],"An AI can treat a digital image as a numerical structure, because every pixel uses a number to describe its color.",[2777],"pixel",{"id":2779,"data":2780,"type":67,"version":24,"maxContentLevel":27},"179dafc9-67c0-487c-9f5f-8057a24d068e",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2781,"multiChoiceCorrect":2783,"multiChoiceIncorrect":2787,"multiChoiceMultiSelect":21,"multiChoiceRevealAnswerOption":6},[2782],"Which of these statements are true of a Generative Adversarial Network (GAN)?",[2784,2785,2786],"The Generator produces AI images","The Discriminator guesses if those images are real","If the Generator fools the Discriminator, the Generator gets positive feedback",[2788,2789,2790],"The Discriminator produces AI images","The Generator guesses if those images are real","If the Discriminator fools the Generator, the Generator gets positive feedback",{"id":2792,"data":2793,"type":24,"maxContentLevel":27,"version":24},"b814488c-9dee-4032-b801-1163055c7e50",{"type":24,"contentRole":42,"markdownContent":2794,"audioMediaId":2795},"The applications of Text-to-Image models are extremely vast and varied.\n\nJust like text generation from an LLM, these images lack the emotional depth and conceptual complexity that you would find in human-made artwork. But if you're willing to overlook that fact...\n\nWith the right choice of prompt, you can use them to generate digital paintings, product blueprints, marketing materials, greetings cards, fashion designs, game assets, memes, and more.\n\n![Graph](image://3713cf3e-0914-4b75-8be0-f8cc99e39e2d \"An octopus loading up Kinnu?\")\n\nWork is even being done to harness these models in medical settings. In the future, a surgeon could potentially describe a patient's symptoms to one of these models, and receive a diagram which helps them to plan the surgery.","39f4f39f-11e8-4a3f-b85d-79528e23a586",{"id":2797,"data":2798,"type":42,"version":24,"maxContentLevel":27,"summaryPage":2800,"introPage":2808,"pages":2814},"5eb155ad-8d95-4df7-9290-1c57393873de",{"type":42,"title":2799},"Other generative AI",{"id":2801,"data":2802,"type":27,"maxContentLevel":27,"version":24},"31fdcc60-2ff6-4a32-9b35-3b64897b49a6",{"type":27,"summary":2803},[2804,2805,2806,2807],"Text-to-Speech models turn written text into spoken words","Text-to-Video models can generate moving visual content","Text-to-Music models can generate songs and melodies","The legality and ethics of training generative AI is uncertain",{"id":2809,"data":2810,"type":29,"maxContentLevel":27,"version":24},"1209dae7-c7d1-4054-bd6d-c5fbac9aa89a",{"type":29,"intro":2811},[2812,2813],"What are the other main types of generative AI?","Why are some of these models being taken to court?",[2815,2828,2845,2850],{"id":2816,"data":2817,"type":24,"maxContentLevel":27,"version":24,"reviews":2820},"2d5b26d8-e44c-4d1e-9d45-f1a0bbbcdcfa",{"type":24,"contentRole":42,"markdownContent":2818,"audioMediaId":2819},"Generative AI isn't only used for text and image generation. There are also lots of other models out there, which can produce a range of content, including speech, videos and music.\n\n**Text-to-Speech models** are able to turn written text into human-like spoken words. They're used all over the place, from voice assistants, to audiobooks, to tools for visually impaired people.\n\nThere are lots of different ways to build these models. But they're generally based on deep learning. If you train a model on thousands of written texts, plus audio files of people reading those texts, it can learn to sound like a real person talking.\n\nAs always, these models aren't aware that they're 'talking'. Just like LLMs, and Text-to-Image models, they treat their output (audio) as a piece of numerical data.\n\n![Graph](image://86ac7ff7-94c2-4b9c-aede-1fb0f74e0d7a \"Audio data. Image via Pexels\")","25d9050f-60cd-4cde-a27d-4cacde24e96a",[2821],{"id":2822,"data":2823,"type":67,"version":24,"maxContentLevel":27},"1209c5ef-ceba-4c0d-88f0-481caa6ccd33",{"type":67,"reviewType":37,"spacingBehaviour":24,"clozeQuestion":2824,"clozeWords":2826},[2825],"Text-to-Speech models turn written text into human-like spoken words.",[2827],"Text-to-Speech",{"id":2829,"data":2830,"type":24,"maxContentLevel":27,"version":24,"reviews":2833},"5e3a611c-5e8c-4a05-8d9c-65f8adade654",{"type":24,"contentRole":42,"markdownContent":2831,"audioMediaId":2832},"**Text-to-Video models** can generate moving visual content after receiving a written prompt. They're essentially just a more advanced and complicated version of a Text-to-Image model, maybe with some Text-to-Speech thrown in if the video is meant to have audio.\n\nThis particular type of generative AI isn't quite as popular as the others. That's mainly because it lacks consistency: even the best models have a tendency to produce slightly strange, uncanny videos.\n\nBut that could potentially change in the next few years, as more advanced models, trained on better datasets, are developed. If that does happen, it could have a major impact on fields like entertainment and advertising.","93c2738d-e6b0-4f41-b0bb-25f24555729d",[2834],{"id":2835,"data":2836,"type":67,"version":24,"maxContentLevel":27},"e1a9af17-2f35-4178-b147-dec7cef9ac9b",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2837,"multiChoiceCorrect":2839,"multiChoiceIncorrect":2841,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[2838],"Why might current Text-to-Video models be less popular than other Gen AI?",[2840],"Their outputs are often uncanny and inconsistent",[2842,2843,2844],"They need more training data than other models","They often lack user-friendly interfaces","They are illegal in a number of countries",{"id":2846,"data":2847,"type":24,"maxContentLevel":27,"version":24},"38f88fa5-ba54-4817-8857-752fd1fd0f62",{"type":24,"contentRole":42,"markdownContent":2848,"audioMediaId":2849},"**Text-to-Music models** are one more type of modern generative AI. They've been trained to generate pieces of music by looking at the patterns and structures of existing compositions.\n\n![Graph](image://07bd2135-c77a-4c32-be85-19063c19fa43 \"Music. Image via Pexels\")\n\nFor example, you could input a prompt like \"generate a calm, relaxing melody,\" and the model would generate a piece of music accordingly.\n\nOne of the most well-known examples of this is a model called Suno AI. But Suno is pretty controversial. In June 2024, it was actually sued by a group of record labels, including Universal, Sony, and Warner Music.\n\nWhy? Because Suno was supposedly trained on a dataset which featured songs from famous artists like Mariah Carey, Green Day, and Michael Jackson. According to the record labels, Suno's use of these songs was breaking copyright laws.","7782d1a5-ebe2-4ba4-823d-749f23b1ba38",{"id":2851,"data":2852,"type":24,"maxContentLevel":27,"version":24,"reviews":2855},"886563d2-050b-4672-9c8f-47592adf43e6",{"type":24,"contentRole":42,"markdownContent":2853,"audioMediaId":2854},"Suno isn't the only AI with a dataset that's come under scrutiny.\n\nA number of artists have also been trying to take Text-to-Image models (like Midjourney) to court, after claiming that these models were trained on their art without getting any prior permission.\n\nIn 2023, a group of authors including George RR Martin, John Grisham, and Jodi Picoult, made similar claims against ChatGPT.\n\nAs things stand, the use of copyrighted content in AI training is still a bit of a gray area. There isn't just the legal question to think about, but the ethical question too. Is it fair to use a creator's work without asking for prior permission?\n\nFor now, let's put a pin in that question. We'll be taking a proper look at AI ethics later in the pathway.","358b38f6-6e3f-435c-b158-8609806ad677",[2856,2870,2877],{"id":333,"data":2857,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":2858,"multiChoiceQuestion":2859,"multiChoiceCorrect":2861,"multiChoiceIncorrect":2862,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":2866,"matchPairsPairs":2867},[330,334,335],[2860],"The company behind which AI model was taken to court by a group of authors in 2023?",[639],[2863,2864,2865],"Midjourney","Stable Diffusion","Claude",[179],[2868],{"left":639,"right":2869,"direction":27},"Taken to court by authors",{"id":2871,"data":2872,"type":67,"version":24,"maxContentLevel":27},"94083e4e-4b30-4d3d-9117-15264bf39bfc",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":2873,"activeRecallAnswers":2875},[2874],"Why might a creator want to take a Gen AI company to court?",[2876],"Their content was used to train an AI without permission",{"id":2878,"data":2879,"type":67,"version":24,"maxContentLevel":27},"b35d7bcb-4131-4c1d-a902-e6a6fec13fe4",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2880,"multiChoiceCorrect":2882,"multiChoiceIncorrect":2884,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[2881],"The company behind which AI model was taken to court by a group of record labels in 2024?",[2883],"Suno AI",[2885,2864,2865],"Gemini",{"id":2887,"data":2888,"type":28,"maxContentLevel":27,"version":42,"orbs":2891},"a369fd7c-f0a0-4586-ba7a-135773b4c55a",{"type":28,"title":2889,"tagline":2890},"Computer Vision","How do computers see?",[2892,2978,3082],{"id":2893,"data":2894,"type":42,"version":42,"maxContentLevel":27,"summaryPage":2895,"introPage":2902,"pages":2908},"1325f8b0-4afe-47fc-950b-6ed73ccd50a2",{"type":42,"title":2889},{"id":2896,"data":2897,"type":27,"maxContentLevel":27,"version":24},"572f4007-717b-4a51-a5e6-d8d45c7ae165",{"type":27,"summary":2898},[2899,2900,2901],"Computers use cameras to 'see', through a process called image acquisition","Cameras capture light on image sensors, which translate the light into numbers","Models are able to interpret images by analyzing these grids of numbers",{"id":2903,"data":2904,"type":29,"maxContentLevel":27,"version":24},"7a92aa65-0f64-4292-8d1d-d3a37a1122e5",{"type":29,"intro":2905},[2906,2907],"How do computers 'see' the world around them?","How do computers analyze and interpret images?",[2909,2914,2931,2944],{"id":2910,"data":2911,"type":24,"maxContentLevel":27,"version":24},"eada0fff-d7b2-4dbb-8be9-58cffff6b96f",{"type":24,"contentRole":42,"markdownContent":2912,"audioMediaId":2913},"It's time to look at another subfield of Artificial Intelligence. Welcome to **Computer Vision**.\n\nComputer Vision is all about a machine's ability to see. In this pathway so far, we've encountered a few examples. Do you remember that model which could tell the difference between photos of cats and photos of dogs? Or the Text-to-Image models which learn by looking at giant datasets of images?\n\nWhat we haven't really talked about, though, is how these models are able to do this. It isn't as though computers have eyes... so how are they able to look at images, and interpret what they see?","5b4c313f-e5bd-44eb-a232-4c6d2e432cbd",{"id":2915,"data":2916,"type":24,"maxContentLevel":27,"version":24,"reviews":2919},"07e9224f-16f6-49c7-b675-93c2f7027557",{"type":24,"contentRole":42,"markdownContent":2917,"audioMediaId":2918},"Computers don't have eyes. But what they *do* have are cameras and sensors.\n\nYou've probably used a digital camera, like the one on your phone, more times than you can count. But have you ever stopped and wondered how that camera works?\n\nWhen light enters the camera, it hits an image sensor. This sensor is divided into tiny squares. Where the light hits each square, it's given a number that describes its color and brightness.\n\n![Graph](image://c802eeae-76ad-4b47-8572-c81bbfb033dd \"Simplified version of image sensor numbers.\")\n\nThese numbers can then be used to reproduce the image on a digital screen. You just need to make sure the pixels on that screen are the same color and brightness as the light that entered the camera.\n\n![Graph](image://8f2be400-3ef8-471a-932c-0f72f948625b \"Pixels. RVB by Zhela912 (CC BY-SA 3.0) \u003Chttps://creativecommons.org/licenses/by-sa/3.0>, via Wikimedia Commons\")\n\nThey can *also* be given to an Artificial Intelligence, which can use these numbers to 'see'. The technical term for this particular process is **image acquisition**.","58b4c945-ac9c-4a71-8b30-b81b6fc5a3c6",[2920],{"id":2921,"data":2922,"type":67,"version":24,"maxContentLevel":27},"1440d187-2668-4fe1-84ea-0f6fe3e71b60",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2923,"multiChoiceCorrect":2925,"multiChoiceIncorrect":2927,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[2924],"When light hits each square in an image sensor, what happens to it?",[2926],"It's assigned a number that describes its color and brightness",[2928,2929,2930],"It's assigned a symbol that describes its color and brightness","It's reflected into the knowledge base","It's reflected into the inference engine",{"id":2932,"data":2933,"type":24,"maxContentLevel":27,"version":42,"reviews":2936},"bb15bb22-e9a4-4610-a26c-b9dc6e543414",{"type":24,"contentRole":42,"markdownContent":2934,"audioMediaId":2935},"So, when we say that an AI is looking at photos of cats and dogs, and learning to tell the difference between them, what it's really doing is analyzing grids of numerical values which represent the color and brightness of individual pixels.\n\nScientists call this particular process **image interpretation**. It's the part which sets an AI apart from just taking that photo on your phone.\n\nThis interpretation is usually done by a specialized type of neural network called a Convolutional Neural Network. We'll look at those in a lot more detail next time – for now, all you really need to know is that these networks are great at learning to interpret images.","e8ff3101-17b1-4167-a4df-58afa7045c28",[2937],{"id":2938,"data":2939,"type":67,"version":24,"maxContentLevel":27},"7c094dc2-c56f-4f5b-962d-3b204218d950",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":2940,"activeRecallAnswers":2942},[2941],"When we say that an AI is 'looking' at an image, what's it really doing?",[2943],"Analyzing grids of numerical values which represent color and brightness",{"id":2945,"data":2946,"type":24,"maxContentLevel":27,"version":24,"reviews":2949},"f4daa629-9814-42ee-942a-d60d2c8b5cb4",{"type":24,"contentRole":42,"markdownContent":2947,"audioMediaId":2948},"Convolutional Neural Networks can be taught to perform a few different types of image interpretation.\n\n**Image classification** involves looking at an image, and working out what it's actually showing. This often includes putting labels around the objects that appear in an image.\n\n**Image segmentation**, on the other hand, involves dividing an image into patches of pixels that represent the exact positions of the objects that appear in an image.\n\n![Graph](image://fed57056-11fc-46a8-bbfa-85ba8ca7e61a \"Classification and segmentation.\")\n\nThere are plenty of other examples, but another interesting one is **pose estimation**, which analyzes the poses of humans in photos. For example, in a photo from a security camera, does a person look threatening or not?\n\n![Graph](image://184eab71-7da9-4bc8-8bde-dbfc5d7abe81 \"Pose estimation.\")","30e8372b-c630-4b28-a15b-9c3c0f33b083",[2950,2967],{"id":2951,"data":2952,"type":67,"version":24,"maxContentLevel":27},"545befe9-852f-464c-a51b-620dc8b0cf2e",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":2953,"matchPairsPairs":2954,"matchPairsShowExamples":6},[179],[2955,2958,2961,2964],{"left":2956,"right":2957,"direction":27},"Image acquisition","Converting light into grids of numbers",{"left":2959,"right":2960,"direction":27},"Image interpretation","Analyzing grids of numbers",{"left":2962,"right":2963,"direction":27},"Image classification","Identifying the type of object in an image",{"left":2965,"right":2966,"direction":27},"Image segmentation","Identifying the position of objects in an image",{"id":2968,"data":2969,"type":67,"version":24,"maxContentLevel":27},"bac7f34c-f22f-408b-9ba1-2c4910871173",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":2970,"multiChoiceCorrect":2972,"multiChoiceIncorrect":2974,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[2971],"What do we call it when an AI interprets the postures of people in images?",[2973],"Pose estimation",[2975,2976,2977],"Posture analysis","Body interpretation","Gesture reading",{"id":2979,"data":2980,"type":42,"version":42,"maxContentLevel":27,"summaryPage":2982,"introPage":2989,"pages":2995},"c94e135c-c105-4607-a96d-86ffdeb5bddd",{"type":42,"title":2981},"Convolutional Neural Networks",{"id":2983,"data":2984,"type":27,"maxContentLevel":27,"version":24},"5d60dcc6-d1c0-4044-88dc-7ca8d99f0f94",{"type":27,"summary":2985},[2986,2987,2988],"CNNs use kernels to scan images for patterns, like corners and edges","Kernels generate a feature map which show where these patterns are found","This feature map can then be interpreted by the AI",{"id":2990,"data":2991,"type":29,"maxContentLevel":27,"version":24},"5b4e0b1c-f5fc-4436-91ef-b0a6a915eb52",{"type":29,"intro":2992},[2993,2994],"Why are Convolutional Neural Networks so good at analyzing images?","What are kernels and feature maps, and why are they so important?",[2996,3017,3039,3044],{"id":2997,"data":2998,"type":24,"maxContentLevel":27,"version":24,"reviews":3001},"8dbe6a4b-e552-44a3-afb4-e766c6d37708",{"type":24,"contentRole":42,"markdownContent":2999,"audioMediaId":3000},"Last time, we mentioned **Convolutional Neural Networks (CNNs)** – a special type of neural network which a lot of the best computer vision systems are based on. Now, we're going to take a look at how these networks actually work.\n\nA CNN is defined by some special hidden layers, which each use something called a **kernel**. This is basically just a tiny filter that can slide back and forth across a digital image, examining little patches of pixels as it goes.\n\nIn the diagram below, you can see how a kernel might examine a series of 3x3 patches as it slides from position 1, to position 2, to position 3.\n\n![Graph](image://676a840d-aa13-41fd-b443-958816e1cb60 \"Kernel sliding across pixels.\")","231aa0ec-a2dc-47b5-9206-1e38f2ecaaba",[3002],{"id":1843,"data":3003,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":3004,"multiChoiceQuestion":3005,"multiChoiceCorrect":3007,"multiChoiceIncorrect":3009,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":3013,"matchPairsPairs":3014},[1838,1841,1842],[3006],"A CNN uses tiny filters that can slide across images, examining patches of pixels. What are these filters called?",[3008],"Kernels",[3010,3011,3012],"Filters","Sliders","Patches",[179],[3015],{"left":3008,"right":3016,"direction":27},"Filters that slide across images",{"id":3018,"data":3019,"type":24,"maxContentLevel":27,"version":24,"reviews":3022},"0b2666ff-4169-4997-abcc-e582ccd98521",{"type":24,"contentRole":42,"markdownContent":3020,"audioMediaId":3021},"Importantly, a kernel will always be trained to look for a particular pattern as it slides across an image. For example, a kernel might be trained to look for a corner, or a dot, or maybe a pattern like the one you can see below.\n\n![Graph](image://9684c940-64d7-44e0-b1fd-e8a61468e0ff \"3x3 pattern.\")\n\nFor every patch of pixels it slides across, the kernel will output a numerical score. This score describes how closely that particular patch of pixels matched the pattern it was looking for.\n\n![Graph](image://22231cd7-325b-47d2-8866-57cffa23fd3d \"Outputting numerical scores.\")\n\nBy the time it's scanned every patch of pixels, we'll have a sheet of numerical scores. This sheet is like a map of the image. Where the numbers are high, we know that the pattern occurred.\n\n![Graph](image://1f61a22f-37a3-48e6-b5f6-5febe41b73e4 \"Sheet of numerical scores.\")","82524c37-4acb-4222-8d94-c254d8684e15",[3023,3032],{"id":3024,"data":3025,"type":67,"version":24,"maxContentLevel":27},"4f52e865-4bda-47ac-8a9f-482f9fc7c7e8",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":3026,"binaryCorrect":3028,"binaryIncorrect":3030},[3027],"As a kernel slides across an image, what is it looking for?",[3029],"A particular pattern, like a corner or a dot",[3031],"Signs of damage or corruption",{"id":3033,"data":3034,"type":67,"version":24,"maxContentLevel":27},"2a3a56b8-81e1-49cd-90cf-4c1d4086ddb1",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":3035,"activeRecallAnswers":3037},[3036],"A kernel will output a numerical score for every patch of pixels in an image. What do these scores represent?",[3038],"How closely each patch of pixels matches up to the pattern that this kernel was looking for",{"id":3040,"data":3041,"type":24,"maxContentLevel":27,"version":24},"e0a7b932-8df2-420d-bb21-124e410b4d51",{"type":24,"contentRole":42,"markdownContent":3042,"audioMediaId":3043},"In a convolutional neural network, this kernel-scanning doesn't only happen once.\n\nEach hidden layer will have a kernel, which is trained to look for its own particular pattern. Every time this happens, we end up with another sheet of numerical scores.\n\nTypically, the kernels in the first few layers will search for relatively simple patterns, like edges, corners and dots. As we get deeper, the kernels search for complex patterns, like shapes or textures. Some might even search for specific objects, like faces, cars, or buildings.","7910c94b-57ef-4833-9cf1-6e2ec88c68e4",{"id":3045,"data":3046,"type":24,"maxContentLevel":27,"version":42,"reviews":3049},"fb7353fd-4f6b-47f2-b935-d85e3d895ce2",{"type":24,"contentRole":42,"markdownContent":3047,"audioMediaId":3048},"By the time all these kernels have scanned the image, we might have thousands of sheets of scores. Together, these become a **feature map** – a large and complex numerical model which represents that image as a whole.\n\n![Graph](image://95b6bbd1-bffb-4038-aff8-ee0546479c05 \"A simplified feature map.\")\n\nIn other words, a CNN is just a way to turn all that intangible data in a real-life image into an interpretable model that computers are able to work with.\n\nCNNs aren't the only way to go about this, but they're a really popular approach to Computer Vision. These models can learn to use their kernels in all kinds of contexts, like recognizing faces, interpreting medical scans, reading handwritten documents, and more.","6f5ee16c-dff4-41f8-aa24-56890965817e",[3050,3065],{"id":713,"data":3051,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":3052,"multiChoiceQuestion":3053,"multiChoiceCorrect":3055,"multiChoiceIncorrect":3057,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":3061,"matchPairsPairs":3062},[709,712,714],[3054],"By the time all the kernels have scanned an image, we'll have thousands of sheets of numbers. Together, they form a numerical model known as what?",[3056],"Feature map",[3058,3059,3060],"Feature graph","Image map","Image graph",[179],[3063],{"left":3056,"right":3064,"direction":27},"Numerical model of image",{"id":3066,"data":3067,"type":67,"version":24,"maxContentLevel":27},"05bacf29-6dd3-4ee0-8874-6a0da84208c1",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":3068,"matchPairsPairs":3070,"matchPairsShowExamples":6},[3069],"In a CNN, different layers often search for different things. Match the pairs below:",[3071,3074,3077,3080],{"left":3072,"right":3073,"direction":27},"The first few layers","Simple patterns, like corners and edges",{"left":3075,"right":3076,"direction":27},"The middle layers","Complex patterns, like shapes or textures",{"left":3078,"right":3079,"direction":27},"The last few layers","Specific objects, like faces or cars",{"left":292,"right":3081,"direction":27},"Specific sounds, like laughter or speech",{"id":3083,"data":3084,"type":42,"version":24,"maxContentLevel":27,"summaryPage":3086,"introPage":3093,"pages":3099},"3940008f-d068-4dff-a9d4-0a380d57fc6a",{"type":42,"title":3085},"Other types of perception",{"id":3087,"data":3088,"type":27,"maxContentLevel":27,"version":24},"f3035d23-c2a5-48bf-a7a6-faef40402200",{"type":27,"summary":3089},[3090,3091,3092],"Computers can hear and smell using machine listening and machine olfaction","They can also use machine taste sytems, and machine touch systems","LiDAR technology uses lasers to create a 3D map of an AI's surroundings",{"id":3094,"data":3095,"type":29,"maxContentLevel":27,"version":24},"743c7498-dd8d-4103-a5fb-d8cbae336b2b",{"type":29,"intro":3096},[3097,3098],"Do computers have the ability to hear and smell?","Do computers have the ability to taste and touch?",[3100,3116,3131,3155],{"id":3101,"data":3102,"type":24,"maxContentLevel":27,"version":24,"reviews":3105},"d2b902ee-5a29-4bf1-b16a-5c2ce721d682",{"type":24,"contentRole":42,"markdownContent":3103,"audioMediaId":3104},"Along with vision, humans have a number of other senses. We can hear things, smell things, taste things, touch things. But can computers do the same?\n\n**Machine Listening** is a catch-all term for computer hearing systems. These systems are based on the same principles as vision, only a microphone sensor is used in place of a camera.\n\n**Machine Olfaction** is the term for smelling systems, which use sensors to detect airborne chemicals. We sometimes call these 'electronic noses' – one model developed in Sydney, Australia, can tell the difference between whiskies just by giving each one a sniff.\n\n![Graph](image://9820fc1e-a5a7-40d4-b7a7-0f59adb182a7 \"Whisky. By Guinnog (CC BY-SA 3.0) \u003Chttps://creativecommons.org/licenses/by-sa/3.0>, via Wikimedia Commons\")","2a16578c-f157-4505-b313-57e92da82bc9",[3106],{"id":3107,"data":3108,"type":67,"version":24,"maxContentLevel":27},"87ea692e-8d7d-4722-9761-60d69fce89cb",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":3109,"multiChoiceCorrect":3111,"multiChoiceIncorrect":3113,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[3110],"An 'electronic nose' in Sydney, Australia, was trained to do what?",[3112],"Smell the difference between whiskies",[3114,3115,173],"Detect explosives","Smell the difference between perfumes",{"id":3117,"data":3118,"type":24,"maxContentLevel":27,"version":24,"reviews":3121},"f7733ffa-90e0-48ac-ba7e-157ee91e1f05",{"type":24,"contentRole":42,"markdownContent":3119,"audioMediaId":3120},"As well as electronic noses, scientists have also developed electronic tongues. We call this field **Machine Taste** – where electronic noses detect airborne chemicals, these tongues detect chemicals in solids and liquids instead.\n\nLast but not least, we have **Machine Touch**. There are a few different ways to go about this. Some researchers have developed whisker-like sensors that measure pressure at the end of each tip.\n\nMore complex approaches use electronic skin, which can detect patterns of pressure over wider areas thanks to an embedded array of sensors. The most cutting-edge examples of electronic skin can even use their sense of touch to measure an object's temperature.","cd986a92-2165-4e26-83f4-8e555ec5ecbb",[3122],{"id":3123,"data":3124,"type":67,"version":24,"maxContentLevel":27},"5154801d-70b8-423c-b859-d7165a2510b0",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":3125,"multiChoiceCorrect":3127,"multiChoiceIncorrect":3128,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[3126],"The most cutting edge examples of electronic skin can be used to detect what?",[157],[3129,3130,158],"Patterns of pressure","Patterns of temperature",{"id":3132,"data":3133,"type":24,"maxContentLevel":27,"version":24,"reviews":3136},"66923bd2-0f34-45be-badd-04dae7a3e90f",{"type":24,"contentRole":42,"markdownContent":3134,"audioMediaId":3135},"All of these different modes of perception have different real-world uses.\n\nAs we've already seen, Computer Vision is great at sorting through images. It's also a useful tool in security systems, with its ability to identify human intruders automatically.\n\n![Graph](image://cd91c102-862d-4f13-87b2-874d70ddeac1 \"Security system. By Sanderflight (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")\n\nMachine Listening can also be used for security, as it automatically reacts to loud or suspicious sounds. It's also important for Natural Language Processing. If you want an AI to understand a voice command, it needs to be able to hear it.\n\nMachine Olfaction can detect dangerous chemicals, like carbon monoxide, in the environment. Machine Taste can be used in food testing, checking for signs of contamination. And Machine Touch is great for factory robots – a sense of touch makes it easier to handle items.\n\n![Graph](image://ba6fd282-2953-4470-8601-4262effb3939 \"Factory robot. Image: (Public domain), via Wikimedia Commons\")","d873872f-31c3-4e15-84de-2fd3b5e21ef6",[3137],{"id":3138,"data":3139,"type":67,"version":24,"maxContentLevel":27},"0014789e-6859-4d98-95eb-3b801c4a8499",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":3140,"matchPairsPairs":3142,"matchPairsShowExamples":6},[3141],"What is a real-life use for each of these?",[3143,3146,3149,3152],{"left":3144,"right":3145,"direction":27},"Machine Listening","Security systems",{"left":3147,"right":3148,"direction":27},"Machine Olfaction","Carbon monoxide sensors",{"left":3150,"right":3151,"direction":27},"Machine Taste","Food contamination testing",{"left":3153,"right":3154,"direction":27},"Machine Touch","Factory robots",{"id":3156,"data":3157,"type":24,"maxContentLevel":27,"version":24,"reviews":3160},"b4f61c54-92cd-4e70-a51f-223ff09f6054",{"type":24,"contentRole":42,"markdownContent":3158,"audioMediaId":3159},"In addition to all these human-like senses, some AI models can also perceive the world around them using methods that humans aren't capable of.\n\n**LiDAR** (which stands for Light Detection and Ranging) is a technology that uses lasers to generate 3D maps of its surroundings.\n\n![Graph](image://1fa1a671-0758-4276-a2a7-5304688e6e7a \"3D LiDAR map. By Ephramac (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")\n\nIt works by sending out laser pulses, then measuring how long it takes for these lasers to bounce back after hitting an object. It's great for something like an AI-driven car, which needs to keep track of moving objects all around it.\n\nBasically, there are lots of different ways for an AI to perceive its surroundings. But they all come down to a similar principle: some kind of sensor will receive an input, then translate this input into a digital version. This digital version is what the AI is able to work with.","7f052560-f46c-454d-904d-59138cda22ac",[3161,3176,3183],{"id":2585,"data":3162,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":3163,"multiChoiceQuestion":3164,"multiChoiceCorrect":3166,"multiChoiceIncorrect":3168,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":3172,"matchPairsPairs":3173},[2583,2584,2580],[3165],"What technology uses lasers to generate 3D maps?",[3167],"LiDAR",[3169,3170,3171],"RADAR","SONAR","HiDAR",[179],[3174],{"left":3167,"right":3175,"direction":27},"Tool for generating 3D maps",{"id":3177,"data":3178,"type":67,"version":24,"maxContentLevel":27},"123a1ffb-4b40-4f71-8df8-0a7fd85bbe5f",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":3179,"activeRecallAnswers":3181},[3180],"What does LiDAR stand for?",[3182],"Light Detection and Ranging",{"id":3184,"data":3185,"type":67,"version":24,"maxContentLevel":27},"61e1fe32-188b-412b-9403-850f065ddf24",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":3186,"matchPairsPairs":3188,"matchPairsShowExamples":6},[3187],"What are each of these designed to interpret?",[3189,3191,3193,3195],{"left":3144,"right":3190,"direction":27},"Audio inputs",{"left":3147,"right":3192,"direction":27},"Chemicals in the air",{"left":3150,"right":3194,"direction":27},"Chemicals in liquids and solids",{"left":3153,"right":3129,"direction":27},{"id":3197,"data":3198,"type":28,"maxContentLevel":37,"version":3201,"orbs":3202},"dc411110-3c5a-4160-85e9-31b08d7842b1",{"type":28,"title":3199,"tagline":3200},"Robotics","How do computers move?",5,[3203,3282,3407],{"id":3204,"data":3205,"type":42,"version":24,"maxContentLevel":27,"summaryPage":3207,"introPage":3214,"pages":3220},"35babc1b-3e20-4801-a3bf-87b29b8f519c",{"type":42,"title":3206},"What is a robot?",{"id":3208,"data":3209,"type":27,"maxContentLevel":27,"version":24},"72273737-7581-46a9-b457-aa720ad9c202",{"type":27,"summary":3210},[3211,3212,3213],"Robots are physical machines that perform tasks impacting the physical world","Robots like NASA's Perseverance rover are exploring the surface of Mars","Robots make some people nervous, in a phenomenon known as the Frankenstein Complex",{"id":3215,"data":3216,"type":29,"maxContentLevel":27,"version":24},"60f60b87-f0cc-4c39-a3ac-0d9f1c4fefc4",{"type":29,"intro":3217},[3218,3219],"What is the definition of a robot?","Is it normal to feel uneasy around robots?",[3221,3226,3239,3256],{"id":3222,"data":3223,"type":24,"maxContentLevel":27,"version":24},"041601b1-c59b-4027-9488-102d58b192dd",{"type":24,"contentRole":42,"markdownContent":3224,"audioMediaId":3225},"It probably feels like a while ago now. But right at the start of this entire pathway, we talked about Talos – that automated giant, made entirely of bronze, which appeared in Greek mythology.\n\n![Graph](image://1709b7e2-6a74-4ca1-8a57-d5c90365c060 \"An AI generated image.\")\n\nThere's something quite distinct about Talos. Something that really sets him apart from most of the real-life AI models we've discussed in this pathway so far. Models like AlphaGo, and ChatGPT, and Midjourney, and Logic Theorist, and so on.\n\nTalos has a physical body. He can pick up stones, he can walk about, he can tangibly interact with the world. In other words, Talos is a **robot**.\n\nAt least, he would be, if he'd actually existed.","9e4e1e12-2a6c-43ea-a4e9-0070072e0754",{"id":3227,"data":3228,"type":24,"maxContentLevel":27,"version":24,"reviews":3231},"f7777093-f5a1-4a75-9d5e-9c98f5591dc5",{"type":24,"contentRole":42,"markdownContent":3229,"audioMediaId":3230},"A modern robot might be defined as follows: a physical machine, performing physical tasks, which impact the physical world. Another popular term is **embodied AI** – that's an AI model with some kind of physical body.\n\nAs we said, a lot of modern models don't fit this definition. ChatGPT lacks a physical body. There's no way for it to physically type, or to use a pen and paper.\n\nBut that doesn't mean that robots don't exist.\n\nJust think of NASA’s Perseverance rover. It's currently using an AI system to physically explore and physically interact with the surface of the planet Mars. Many factories are now using robots too, to assemble products and transport items around.\n\n![Graph](image://52db0336-3711-491d-abf9-bb6f2cd9f680 \"Perseverance rover. Image: (Public domain), via Wikimedia Commons\")","372f1413-33e9-4c50-8fb1-80e44ba2dfdd",[3232],{"id":3233,"data":3234,"type":67,"version":24,"maxContentLevel":27},"c918d6d9-dd02-45a2-8631-49efee144726",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":3235,"activeRecallAnswers":3237},[3236],"What's the modern definition of a robot?",[3238],"A physical machine, performing physical tasks, which impact the physical world",{"id":3240,"data":3241,"type":24,"maxContentLevel":27,"version":24,"reviews":3244},"625e7c27-0e3a-4264-8260-35c48f6fe939",{"type":24,"contentRole":42,"markdownContent":3242,"audioMediaId":3243},"Factory robots, and other similar machines, are a relatively modern invention. And if you talk to people, these robots actually make a lot of people nervous.\n\nAccording to Isaac Asimov – a celebrated science fiction writer, who wrote extensively about robots during the twentieth century – humans are intrinsically frightened of intelligent machines. We fear that robots will rise up and attack us, something which a disembodied AI (like ChatGPT) wouldn't possibly be able to do.\n\nAsimov called this the **Frankenstein Complex**: the fear that robots will turn against us, just like the monster in *Frankenstein* turned against his creator.","519602b1-4334-4a71-b7d3-fd07c1fa0bf1",[3245],{"id":3246,"data":3247,"type":67,"version":24,"maxContentLevel":27},"793f6954-c698-4502-971f-c6be2cb98512",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":3248,"multiChoiceCorrect":3250,"multiChoiceIncorrect":3252,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[3249],"Isaac Asimov theorized that humans are intrinsically frightened of robots. What did he call this fear?",[3251],"Frankenstein Complex",[3253,3254,3255],"Talos Complex","Frankenstein Syndrome","Talos Syndrome",{"id":3257,"data":3258,"type":24,"maxContentLevel":27,"version":24,"reviews":3261},"32d68599-475a-4285-8fe5-be7baf8ec6f1",{"type":24,"contentRole":42,"markdownContent":3259,"audioMediaId":3260},"Of course, the idea of a robot turning against us isn't really realistic. At least, not with the robots which are currently being used.\n\nEmbodied or otherwise, these are still Artificial Narrow Intelligence models. All they can do is perform the tasks they were originally designed for, like fitting two pieces of metal together, or moving a box from point A to point B. They can't make the leap to 'vengeance' or 'violence' – that's not what they were built to do.\n\nIn other words, it's best to view robotics as an exciting technology, not a worrying one. As things stand, the idea of a robot uprising isn't a valid concern.\n\n![Graph](image://6984d3f6-a543-427b-a990-aa883c5285a9 \"Robot non-vengefully sawing. By Luka Peternel (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")","46c9fb98-ba7c-4af7-ab49-e2056cb58a33",[3262,3273],{"id":3263,"data":3264,"type":67,"version":24,"maxContentLevel":27},"1b10426c-347c-4f27-a6b8-618b998313ea",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":3265,"multiChoiceCorrect":3267,"multiChoiceIncorrect":3269,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[3266],"Which of these is an alternative term for a robot?",[3268],"Embodied AI",[3270,3271,3272],"Corporeal AI","Externalized AI","Material AI",{"id":3274,"data":3275,"type":67,"version":24,"maxContentLevel":27},"09422361-4ec8-43e1-b7b9-56522a681e0f",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":3276,"multiChoiceCorrect":3278,"multiChoiceIncorrect":3280,"multiChoiceMultiSelect":21,"multiChoiceRevealAnswerOption":6},[3277],"Which of these are real-life examples of robots?",[3279,3154],"NASA's Perseverance rover",[3281,639,2863],"Talos",{"id":3283,"data":3284,"type":42,"version":3201,"maxContentLevel":37,"summaryPage":3286,"introPage":3293,"pages":3299},"c0f4938d-92c0-4c0e-921c-c61e7805610d",{"type":42,"title":3285},"Robotic bodies",{"id":3287,"data":3288,"type":27,"maxContentLevel":27,"version":24},"2366ac52-3383-49e3-80a8-45723271085e",{"type":27,"summary":3289},[3290,3291,3292],"Humanoid robots look like humans, while non-humanoid robots can take on various forms","Humanoid robots are mainly used for research and entertainment","Non-humanoid robots include industrial robots and Automated Guided Vehicles",{"id":3294,"data":3295,"type":29,"maxContentLevel":27,"version":42},"1ec01c5e-9344-469a-b925-156b3899104c",{"type":29,"intro":3296},[3297,3298],"What are humanoid robots primarily used for?","What are non-humanoid robots used for?",[3300,3332,3353,3370],{"id":3301,"data":3302,"type":24,"maxContentLevel":37,"version":42,"reviews":3305},"cf962ff8-6d5e-407a-a158-24658115f45a",{"type":24,"contentRole":42,"markdownContent":3303,"audioMediaId":3304},"As we talked about last time, a robot is defined by its physical body. That's what sets it apart from disembodied AIs like ChatGPT or AlphaGo. But we haven't talked about how these bodies actually work.\n\nAn **effector** is the part of a robot's body that actually performs an action. For example, it might have a gripper (for picking up objects), a drill (for cutting through metal), or a wheel (for moving around).\n\nAn **actuator**, on the other hand, provides the physical force that allows an effector to move. For example, it might use an electric motor to turn its wheels, or a pneumatic (compressed air) system that forces its gripper to close.","7aa8f38b-8597-4fed-969d-932474af21d9",[3306,3321],{"id":2583,"data":3307,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":3308,"multiChoiceQuestion":3309,"multiChoiceCorrect":3311,"multiChoiceIncorrect":3313,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":3317,"matchPairsPairs":3318},[2584,2580,2585],[3310],"What do we call the part of a robot that performs an action?",[3312],"Effector",[3314,3315,3316],"Actuator","Manipulator","Controller",[179],[3319],{"left":3312,"right":3320,"direction":27},"Performs an action",{"id":2584,"data":3322,"type":67,"version":42,"maxContentLevel":37},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":3323,"multiChoiceQuestion":3324,"multiChoiceCorrect":3326,"multiChoiceIncorrect":3327,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":3328,"matchPairsPairs":3329},[2583,2580,2585],[3325],"What do we call the part of a robot that provides physical force?",[3314],[3312,3315,3316],[179],[3330],{"left":3314,"right":3331,"direction":27},"Provides physical force",{"id":3333,"data":3334,"type":24,"maxContentLevel":27,"version":24,"reviews":3337},"e616603c-041d-4045-ae97-711178b4aacc",{"type":24,"contentRole":42,"markdownContent":3335,"audioMediaId":3336},"So, a robotic body is just a set of effectors, powered by a set of actuators. But depending on the nature of the robot, these parts can be assembled in totally different ways.\n\nOn the one hand, we have **humanoid robots**. These are mechanical bodies which are put together to look like real humans. One famous example is NASA's Valkyrie, which basically looks like an actual person wrapped up in a high-tech suit.\n\n![Graph](image://8aa27743-07c6-4a6d-950a-7c63e3a408a6 \"Valkyrie. Image: (Public domain), via Wikimedia Commons\")\n\nOn the other hand, we have **non-humanoid robots**. These are mechanical bodies which don't look anything like us. Think of the Perseverance rover, or a self-flying drone, or a floor-mounted arm like the ones you might find in a factory.\n\n![Graph](image://8f536a1f-420e-47a5-ba71-6edf73cd0e3e \"Floor-mounted arms. Image: (Public domain), via Wikimedia Commons\")","103bcd56-9d86-4912-ba23-c9cf7c6e2ac0",[3338],{"id":3339,"data":3340,"type":67,"version":24,"maxContentLevel":27},"101c3adf-4e7d-4521-9079-fb2825ba8f64",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":3341,"matchPairsPairs":3342,"matchPairsShowExamples":6},[179],[3343,3345,3347,3350],{"left":3312,"right":3344,"direction":27},"Gripping device",{"left":3314,"right":3346,"direction":27},"Electric motor",{"left":3348,"right":3349,"direction":27},"Humanoid robot","NASA's Valkyrie",{"left":3351,"right":3352,"direction":27},"Non-humanoid robot","Self-flying drone",{"id":3354,"data":3355,"type":24,"maxContentLevel":27,"version":37,"reviews":3358},"da0f9292-c00d-4228-a7bf-c54b8e289499",{"type":24,"contentRole":42,"markdownContent":3356,"audioMediaId":3357},"It has to be said: humanoid robots don't currently have many useful, real-world uses. For the time being, you won't see one working in a factory, or delivering your morning mail.\n\nThey're generally built for research purposes. One well-known example is the iCub robot, which was built in the style of a human child, then encouraged to learn some child-like behaviors, like crawling, playing, and talking.\n\n![Graph](image://616fa92f-4923-45f4-8acf-6fc280ff74c3 \"iCub robot. Image: Xavier Caré / Wikimedia Commons / CC-BY-SA, via Wikimedia Commons\")\n\nHumanoid robots are also built for entertainment. In 2016, Hanson Robotics launched Sophia – a humanoid robot which caught the attention of the world. Sophia is essentially just a chatbot with effectors for facial expressions, but she's given a number of televised interviews, and even spoken at important events.\n\n![Graph](image://49630ac1-0b89-47d7-95c1-3a77daa4258e \"Sophia robot. Image: (cropped) by ITU Pictures (CC BY 2.0) \u003Chttps://creativecommons.org/licenses/by/2.0>, via Wikimedia Commons\")","8d68a9c4-e694-4f76-8549-cec107259ea0",[3359],{"id":3360,"data":3361,"type":67,"version":27,"maxContentLevel":27},"f8f05401-dc0c-4695-8591-62d8c4823903",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":3362,"multiChoiceCorrect":3364,"multiChoiceIncorrect":3367,"multiChoiceMultiSelect":21,"multiChoiceRevealAnswerOption":6},[3363],"As things stand, what are the two main purposes of humanoid robots?",[3365,3366],"Research studies","General entertainment",[3368,3369],"Factory work","Home assistance",{"id":3371,"data":3372,"type":24,"maxContentLevel":37,"version":27,"reviews":3375},"abdb6785-7b22-4ac4-82c9-58041e86ee9e",{"type":24,"contentRole":42,"markdownContent":3373,"audioMediaId":3374},"While humanoid robots are mainly just used for research and entertainment, non-humanoid robots have lots of real-world uses.\n\nAn **industrial robot** is a class of robot which is generally used in factories. These often take the form of robotic arms, which use Computer Vision to look at objects, before using their effectors to pick them up and manipulate them.\n\nIn general, these robots are used for long, repetitive tasks which would be difficult for a human to perform. For example, if a factory needed to hammer a million bolts into place, it might be better to give that work to a robot, rather than a human who might tire out.\n\nAnother example of a non-humanoid robot is an **automated guided vehicle (AGV)**. It uses vision or LiDAR to navigate a space, like a factory building, where it carefully transports goods from place to place.\n\n![Graph](image://8df4d5a5-7936-486c-8456-2f5f5c9eeb3e \"AGV. By Raimond Spekking (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")","dd4d5732-cc6b-44e7-837d-1a19a6bea41f",[3376,3393,3400],{"id":3377,"data":3378,"type":67,"version":24,"maxContentLevel":27},"1f8a6fe4-29e7-4ce8-9c23-5a6337f42083",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":3379,"matchPairsPairs":3380,"matchPairsShowExamples":6},[179],[3381,3384,3387,3390],{"left":3382,"right":3383,"direction":27},"iCub robot","A research project which learns like a human child",{"left":3385,"right":3386,"direction":27},"Sophia robot","An AI chatbot with a human-like body and face",{"left":3388,"right":3389,"direction":27},"Industrial robot","A class of robot used in factories",{"left":3391,"right":3392,"direction":27},"AGV","A robot which transports goods",{"id":3394,"data":3395,"type":67,"version":42,"maxContentLevel":37},"4acaad89-a14c-4ada-a9d7-9351ac795875",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":3396,"activeRecallAnswers":3398},[3397],"In the context of robotics, what does AGV stand for?",[3399],"Automated Guided Vehicle",{"id":3401,"data":3402,"type":67,"version":42,"maxContentLevel":27},"dd6d1d49-dc08-4b28-a657-5bc694b3ffc0",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":3403,"activeRecallAnswers":3405},[3404],"What type of task is an industrial robot best suited for?",[3406],"Repetitive tasks which might tire a human worker",{"id":3408,"data":3409,"type":42,"version":24,"maxContentLevel":27,"summaryPage":3411,"introPage":3418,"pages":3424},"bd1f0a03-3cfd-4b2f-bf62-f60a57b9e386",{"type":42,"title":3410},"Robotic minds",{"id":3412,"data":3413,"type":27,"maxContentLevel":27,"version":24},"4ab70c38-c176-444d-a824-034399bdbf75",{"type":27,"summary":3414},[3415,3416,3417],"Robots don't always use AI, but the most useful ones usually do","They can use AI for various tasks, like quality assurance or navigation","Building a humanoid robot like Talos would potentially now be possible",{"id":3419,"data":3420,"type":29,"maxContentLevel":27,"version":24},"fe0710fd-12af-4580-9307-6136ede44cb6",{"type":29,"intro":3421},[3422,3423],"What tasks can robots use AI for?","Is building a humanoid robot like Talos now possible?",[3425,3438,3443,3448,3453],{"id":3426,"data":3427,"type":24,"maxContentLevel":27,"version":24,"reviews":3430},"5cef75c7-8776-4904-8260-c2c8c2159f21",{"type":24,"contentRole":42,"markdownContent":3428,"audioMediaId":3429},"It's time to come clean about something. We know this is a pathway about Artificial Intelligence... but a lot of robots don't actually use AI.\n\nConsider a simple, mechanical arm. It's been programmed to saw a piece of wood. When you press a button, that's exactly what it does.\n\nThis arm is a robot: a physical machine, performing a physical task, which impacts the physical world. But does this arm use AI? Not at all. It doesn't need to 'think' or 'make decisions'. It just follows its programming, and moves that saw back and forth, whenever you press the button.\n\nThink of it like this. A robot is a mechanical body. But that body doesn't *necessarily* need a 'mind'.","acf282dc-cc79-4993-9cfc-3dcbfdd0f0ce",[3431],{"id":3432,"data":3433,"type":67,"version":24,"maxContentLevel":27},"338420ac-faf4-414d-80e9-6974c17b0988",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":3434,"multiChoiceCorrect":3436,"multiChoiceIncorrect":3437,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[3435],"True or false: robots are often controlled by AI, but not always.",[1161],[1163],{"id":3439,"data":3440,"type":24,"maxContentLevel":27,"version":24},"b5f40823-1570-4587-842c-5052c91e54b2",{"type":24,"contentRole":42,"markdownContent":3441,"audioMediaId":3442},"So, not all robots use AI. But a lot of the best ones do. Let's consider, for example, an AGV which is designed to pick up trash from the factory floor, then deliver that trash to a bin.\n\nFirst of all, that robot will use Computer Vision to spot any objects on the floor. It will also need an image classification system, which helps it to identify whether those objects actually need to be thrown away.\n\nThis image classification system is probably some kind of neural network, which has been trained on a labeled dataset of trash (waste paper, burned wood, sawdust) and not trash (hammers, buckets, screws).\n\n![Graph](image://f6f5d496-e1ef-4718-8d49-87f08ab847cf \"Trash or not trash?\")","eb0d7802-5dc7-464e-b227-109fa9466945",{"id":3444,"data":3445,"type":24,"maxContentLevel":27,"version":24},"70e6872c-fd6c-4554-b97a-9bac16a49332",{"type":24,"contentRole":42,"markdownContent":3446,"audioMediaId":3447},"Once the AI system has identified a piece of trash on the floor, it will send signals to the robot's actuators. These actuators, in turn, power the robot's effectors. It wheels its way over to the piece of trash, and picks it up with a grabber.\n\nNow, the robot needs to make its way to the bin. But the factory is crowded with human workers walking back and forth.\n\n![Graph](image://0ceb56e3-9fc9-4149-bcda-d27c8d304905 \"Factory floor.\")\n\nLuckily, this robot has a LiDAR system. It starts sending out lasers, which bounce off the humans, and give the robot a dynamic, 3D map of all these moving workers. It uses this map to navigate the space, still sending signals to actuators which cause its wheels to steer left and right as required.","a19b8c55-87e2-40d9-9749-28f465838f88",{"id":3449,"data":3450,"type":24,"maxContentLevel":27,"version":24},"d76ae1f4-b08f-4f6a-ba1c-0294e7de3fce",{"type":24,"contentRole":42,"markdownContent":3451,"audioMediaId":3452},"Eventually, the robot reaches the bin. The AI sends a signal to another actuator, which opens the grabber, and drops the trash into the bin.\n\nThe robot is about to turn around, and search the factory for another piece of trash, when its microphones pick up the sound of a voice: \"Hey, robot, that's enough for today. Can you switch into sleep mode?\"\n\nThe AI takes these words, and uses some Natural Language Processing to translate them into numbers. It recognizes these numbers as a command, and responds by sending out a set of signals which shuts its system down.","996355d8-0cbd-4d6d-ab7d-b92159fcfd24",{"id":3454,"data":3455,"type":24,"maxContentLevel":27,"version":24,"reviews":3458},"2fc17349-4e7c-4e53-9a70-4633b99be41a",{"type":24,"contentRole":42,"markdownContent":3456,"audioMediaId":3457},"That AGV is just one example of a robot using AI.\n\nAnother robot, which was built to weave colorful rugs, might use Generative AI to come up with new designs. Another might use Computer Vision to perform quality assurance on products. Another might use NLP to give reports and updates to workers. There are honestly countless examples. The list goes on and on.\n\nAnd it makes you wonder... if we wanted to build a robot like Talos, would that actually now be possible?\n\nA humanoid robot which walks back and forth, using Computer Vision to watch the ocean. When it sees an object which resembles a boat, it picks up a rock using actuators and effectors, and throws that rock at the object. When the object sinks, the robot continues its patrol.","0b5cd323-90cd-4c2d-9ebc-4fad2dc14d78",[3459],{"id":3460,"data":3461,"type":67,"version":24,"maxContentLevel":27},"98170b63-0885-4e82-8a34-65fad371fc39",{"type":67,"reviewType":177,"spacingBehaviour":24,"matchPairsQuestion":3462,"matchPairsPairs":3464,"matchPairsShowExamples":6},[3463],"How might a factory robot use these different types of AI?",[3465,3468,3470,3472],{"left":3466,"right":3467,"direction":27},"Computer vision","Checking the quality of products",{"left":3167,"right":3469,"direction":27},"Navigating the factory floor",{"left":2077,"right":3471,"direction":27},"Obeying commands from workers",{"left":2510,"right":3473,"direction":27},"Coming up with product designs",{"id":3475,"data":3476,"type":28,"maxContentLevel":27,"version":27,"orbs":3479},"46b5c806-93d0-402f-8473-75006dbc6c35",{"type":28,"title":3477,"tagline":3478},"AI Ethics","Are we doing the right thing?",[3480,3581,3684,3767,3863],{"id":3481,"data":3482,"type":42,"version":42,"maxContentLevel":27,"summaryPage":3484,"introPage":3491,"pages":3497},"4dc6cb10-cd5f-4fcb-b3a6-38c3a1955f53",{"type":42,"title":3483},"Plagiarism",{"id":3485,"data":3486,"type":27,"maxContentLevel":27,"version":24},"17aefdd3-1feb-48c0-81c7-ab9a0bfcb1ca",{"type":27,"summary":3487},[3488,3489,3490],"AI companies have been accused of using copyrighted content for training","Copyrighted content in AI training data is a grey area legally and ethically","Creators argue they should have control over how their art is used",{"id":3492,"data":3493,"type":29,"maxContentLevel":27,"version":24},"202b9afd-a69e-4e93-92dc-72b1f99c8a0d",{"type":29,"intro":3494},[3495,3496],"Is it legal to use copyrighted content for AI training data?","Should creators decide whether or not their content is used?",[3498,3503,3543,3561],{"id":3499,"data":3500,"type":24,"maxContentLevel":27,"version":24},"cb811d92-6b1b-4716-b3c0-2b33a73889b6",{"type":24,"contentRole":42,"markdownContent":3501,"audioMediaId":3502},"In this pathway so far, we've looked at lots of different ways that Artificial Intelligence is changing the modern world. It's never been easier to find patterns in data, to generate content, to automate processes, and so on.\n\nBut here's the thing: a lot of people are actually unhappy with these changes. They've raised questions about the ethics of AI: should we really be using these technologies? Maybe the rise of modern AI will have a negative impact on the world.\n\nOver the course of this tile, we'll be taking a look at these ethical questions in more detail. It's important to stop and do this sometimes. To step back from the hype, and think about AI from a slightly different angle.","f7d4c8ba-f564-483c-a686-6fcfd8f0a2bc",{"id":3504,"data":3505,"type":24,"maxContentLevel":27,"version":24,"reviews":3508},"927611c5-b7f2-494e-84cf-f69d1b106460",{"type":24,"contentRole":42,"markdownContent":3506,"audioMediaId":3507},"The first ethical concern we'll be taking a look at is **ownership**.\n\nWe already touched on this subject earlier, when we learned about Generative AI. Essentially, whenever an AI company trains a model, it requires a lot of data. And sometimes, this data (e.g. artwork, text) is scraped from the internet without asking for permission from the creators.\n\nIn 2023, Getty Images launched a lawsuit against Stability AI. Getty claimed that Stability had used 12 million of their licenced images to train a Text-to-Image model, all without Getty's consent.\n\nSimilar cases have also been made against Large Language Models. The New York Times launched a similar lawsuit against OpenAI, also in 2023. OpenAI supposedly used copyrighted online articles while training ChatGPT.","d73a9f08-e11f-4f5d-aa8a-23c0cf700db5",[3509,3528],{"id":3510,"data":3511,"type":67,"version":24,"maxContentLevel":27},"7e093f96-ae33-4b4f-a183-829d3aa27e73",{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":3512,"multiChoiceQuestion":3516,"multiChoiceCorrect":3518,"multiChoiceIncorrect":3520,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":3524,"matchPairsPairs":3525},[3513,3514,3515],"496cff80-3932-45fd-bdea-6c71d96e8a8b","ab8d6ae2-ce65-43de-a2e9-85cf6a402101","0c028ab0-31b7-4173-8550-0832f427cbcd",[3517],"In 2023, which of these companies launched a lawsuit against Stability AI?",[3519],"Getty Images",[3521,3522,3523],"Adobe","Microsoft","iStock Images",[179],[3526],{"left":3519,"right":3527,"direction":27},"Launched a lawsuit against Stability AI",{"id":3513,"data":3529,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":3530,"multiChoiceQuestion":3531,"multiChoiceCorrect":3533,"multiChoiceIncorrect":3535,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":3539,"matchPairsPairs":3540},[3510,3514,3515],[3532],"In 2023, which of these companies launched a lawsuit against OpenAI?",[3534],"New York Times",[3536,3537,3538],"Washington Post","CNN","Wall Street Journal",[179],[3541],{"left":3534,"right":3542,"direction":27},"Launched a lawsuit against OpenAI",{"id":3544,"data":3545,"type":24,"maxContentLevel":27,"version":42,"reviews":3548},"742c0b60-7bf6-4da2-821a-0cb44348a90e",{"type":24,"contentRole":42,"markdownContent":3546,"audioMediaId":3547},"Is the use of copyrighted content in training data actually against the law? AI is still such a new technology, that it's all a bit of a gray area. Those court cases we mentioned will be pretty important in setting a precedent for the future.\n\nPutting aside the legal question, though, some people argue that this practice just isn't very ethical. If someone produces a painting, or a novel, or any other piece of creative content, shouldn't they get to decide how it's used?\n\nThat's what a lot of creators are saying. In 2023, an artist by the name of Eva Toorenent said this to the BBC: \"If I'm the owner, I should decide what happens to my art.\"","817b143a-3636-4984-adf3-2cc74f6a0f57",[3549],{"id":3550,"data":3551,"type":67,"version":42,"maxContentLevel":27},"41a61463-4159-45ce-a2f1-16b9b606a2fb",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":3552,"activeRecallAnswers":3554,"multiChoiceQuestion":3556,"multiChoiceCorrect":3557,"multiChoiceIncorrect":3558,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[3553],"Is the use of copyrighted content in training data currently against the law?",[3555],"It's a grey area",[3553],[3555],[3519,3559,3560],"Stability AI","BBC",{"id":3562,"data":3563,"type":24,"maxContentLevel":27,"version":24,"reviews":3566},"5f6fc439-28b8-446d-912b-3c7518866fd2",{"type":24,"contentRole":42,"markdownContent":3564,"audioMediaId":3565},"An AI company might make the argument: human creators are constantly taking inspiration from one another. An artist might go to a local gallery, encounter a painting... and if they like it enough, they might shift the way they produce their own works of art.\n\n![Graph](image://95d885a8-5d28-4d7e-8663-cd6395cf2d50 \"Art gallery. Image via Pexels\")\n\nIs it different, then, to show a piece of art to a neural network, and let it be inspired too? If it *is* different, then *why* is it different? Again, it's a bit of a gray area, and you can make arguments for either side.\n\nAt the end of the day, this is all uncharted territory. The AI industry is moving so fast that it's forcing us to think about ownership, inspiration, copyright and consent in ways that never would have crossed our minds in the past.","8bb55576-5c7a-4dd2-9d5c-9d32080c214f",[3567,3574],{"id":3568,"data":3569,"type":67,"version":24,"maxContentLevel":27},"d5d98538-26a8-4b91-b4cc-f2a1e38aa8e6",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":3570,"activeRecallAnswers":3572},[3571],"A company might argue that the use of content to train an AI is no different to what?",[3573],"A human taking inspiration from someone else's work",{"id":3575,"data":3576,"type":67,"version":24,"maxContentLevel":27},"a666cc65-bbee-42d7-a58a-1e2cb25dc7fb",{"type":67,"reviewType":37,"spacingBehaviour":24,"clozeQuestion":3577,"clozeWords":3579},[3578],"In 2023, artist Eva Toorenent told the BBC: \"If I'm the owner, I should decide what happens to my art.\"",[3580],"owner",{"id":3582,"data":3583,"type":42,"version":42,"maxContentLevel":27,"summaryPage":3585,"introPage":3592,"pages":3598},"17a03241-14f3-40b6-8dbc-840094d2140c",{"type":42,"title":3584},"Misinformation",{"id":3586,"data":3587,"type":27,"maxContentLevel":27,"version":24},"76c1f1d3-9846-4470-8b21-4ac23f45f144",{"type":27,"summary":3588},[3589,3590,3591],"Hallucination means that an AI model can spread misinformation without realizing","Hallucination is hard to trace due to a lack of explainability in neural networks","Deepfakes are a type of AI-generated content that make it look like someone did something they never actually did",{"id":3593,"data":3594,"type":29,"maxContentLevel":27,"version":24},"717e6629-2ae2-4035-a17a-7d02264075a6",{"type":29,"intro":3595},[3596,3597],"How do AI models unintentionally spread misinformation?","Why is it so hard to stop this from happening?",[3599,3612,3633,3655],{"id":3600,"data":3601,"type":24,"maxContentLevel":27,"version":42,"reviews":3604},"2ab69ce9-83ff-4018-a51b-bd72f1a5f7ba",{"type":24,"contentRole":42,"markdownContent":3602,"audioMediaId":3603},"Artificial Intelligence has a problem. Over the last few years, it's been repeatedly involved with the spread of **misinformation**.\n\nHere's the thing. AI systems are only as good as the data they're originally trained on. If the data contains any mistakes, or falsehoods, they can be replicated by the AI. Scientists call this GIGO: 'Garbage In, Garbage Out'.\n\nFor example, if an AI was trained on a dataset of random online articles, and a few of those articles happened to claim that the Earth is flat... there's a chance the AI will regurgitate that view.\n\n**Bias** is also a problem. Imagine a model which helps companies to hire new staff. If it's trained on data from previous hires, and those hires favored a certain race or gender, the AI is likely to have those biases too.","e0ff08dc-cd4f-4003-b46e-c1991edcb95d",[3605],{"id":3606,"data":3607,"type":67,"version":24,"maxContentLevel":27},"4558db5e-6d61-497a-abb0-a7c4e1745fa9",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":3608,"activeRecallAnswers":3610},[3609],"Why might an AI model be biased against certain races or genders?",[3611],"Because its training data was biased against those races or genders",{"id":3613,"data":3614,"type":24,"maxContentLevel":27,"version":42,"reviews":3617},"f6845d86-62ab-4f87-b3a1-b0b0134051db",{"type":24,"contentRole":42,"markdownContent":3615,"audioMediaId":3616},"Remember: when an AI produces misinformation, or biased outputs, it has no idea that it's doing it. It has no sense of 'truth' or 'prejudice' – it's just replicating patterns in the data.\n\nExperts call it **hallucination**, and it can happen in all kinds of contexts. From misdiagnosing medical conditions, to referencing sources that never existed, to producing images like the one you can see below.\n\nIn 2022, a chatbot being used on Air Canada's website started promising discounts to passengers. These discounts weren't actually real. The chatbot was just hallucinating.\n\nBut in 2024, a court decided that Air Canada had to honor these discounts. It's put some companies on red alert – it's a risky game to use an AI model if you're held accountable for its mistakes.\n\n![Graph](image://c2d97cdd-0841-463d-82ac-4d1c43394257 \"Air Canada. By Mark Harkin (CC BY 2.0) \u003Chttps://creativecommons.org/licenses/by/2.0>, via Wikimedia Commons\")","771c781c-570f-4891-b167-a5c3889c778e",[3618],{"id":1142,"data":3619,"type":67,"version":42,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":3620,"multiChoiceQuestion":3621,"multiChoiceCorrect":3623,"multiChoiceIncorrect":3625,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":3629,"matchPairsPairs":3630},[1141,1143,1138],[3622],"What do experts call it when an AI produces biased or incorrect outputs?",[3624],"Hallucination",[3626,3627,3628],"Imagination","Aberration","Deepfake",[179],[3631],{"left":3624,"right":3632,"direction":27},"Producing biased or incorrect outputs",{"id":3634,"data":3635,"type":24,"maxContentLevel":27,"version":24,"reviews":3638},"8286da33-7a02-467f-8a02-13ad5f19d6e1",{"type":24,"contentRole":42,"markdownContent":3636,"audioMediaId":3637},"A major problem with hallucination is that it's hard to know where the mistakes and biases even came from.\n\nWhen a neural network 'makes a decision', it will often involve a chain of parameters that might be millions or even billions of steps long. And none of those steps are labeled. There's no way for a human to check where the model veered off course from a 'truthful' output, and found itself spouting misinformation instead.\n\nThat's why a lot of AI experts are now pushing for **explainability**. This means building models with decision making processes that are more visible and easy to understand.\n\nThat's easier said than done. It would require a rethink of our current approach to neural networks. But if it *could* be done, it would make it much easier to track down (and fix) any problems with hallucination.","740bb049-a2f9-44b8-a5cd-8d2a5fd99a9d",[3639,3648],{"id":3640,"data":3641,"type":67,"version":24,"maxContentLevel":27},"aefcf499-93aa-4672-b722-c255fbe8bd62",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":3642,"binaryCorrect":3644,"binaryIncorrect":3646},[3643],"In the context of AI, what is explainability?",[3645],"Visible decision making processes",[3647],"Visible inputs and outputs",{"id":3649,"data":3650,"type":67,"version":24,"maxContentLevel":27},"66da696e-da44-4d83-8645-3377521482f7",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":3651,"activeRecallAnswers":3653},[3652],"Why would better explainability help to combat hallucination?",[3654],"It would make it easier to find and fix whatever caused the error",{"id":3656,"data":3657,"type":24,"maxContentLevel":27,"version":24,"reviews":3660},"d2fb9576-e36d-4f31-a5d7-d2e8be8d141c",{"type":24,"contentRole":42,"markdownContent":3658,"audioMediaId":3659},"With hallucination, AI models can spread misinformation by accident. But they're also a very powerful tool for people who want to spread misinformation on purpose.\n\nYou might have heard of **deepfakes**. These are AI-generated videos and images that make it look like someone is doing something that they never actually did. They’re called deepfakes because they use 'deep' learning models to 'fake' this content in a very realistic way.\n\n![Graph](image://b89541e3-989f-472c-a3dd-e7b7baebdb56 \"The Pope never wore this coat. Image: (Public domain), via Wikimedia Commons\")\n\nFor example, on the eve of an important election, a deepfake might show a famous politician kicking an innocent puppy. If this deepfake went viral, it could potentially be seen by millions of people, and change who those people vote for.\n\nOne thing is certain: in the age of AI, we can no longer trust our own eyes.","4cc45b49-f5a0-451d-a0ef-307845c118af",[3661,3676],{"id":3514,"data":3662,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":3663,"multiChoiceQuestion":3664,"multiChoiceCorrect":3666,"multiChoiceIncorrect":3668,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":3672,"matchPairsPairs":3673},[3510,3513,3515],[3665],"Which of these companies had to pay out to customers after its chatbot started offering non-existent refunds?",[3667],"Air Canada",[3669,3670,3671],"Delta Airlines","British Airways","United Airlines",[179],[3674],{"left":3667,"right":3675,"direction":27},"Used a chatbot that hallucinated refunds",{"id":3677,"data":3678,"type":67,"version":24,"maxContentLevel":27},"f90c67df-2d60-4a9f-9449-e8d1a16b196c",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":3679,"multiChoiceCorrect":3681,"multiChoiceIncorrect":3682,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[3680],"What term describes a deliberately misleading image or video produced by an AI model?",[3628],[3683,3624,3627],"Fakeout",{"id":3685,"data":3686,"type":42,"version":24,"maxContentLevel":27,"summaryPage":3688,"introPage":3695,"pages":3701},"79ab0214-0c2d-4bdf-bc74-7f4f020b99fa",{"type":42,"title":3687},"Weaponization",{"id":3689,"data":3690,"type":27,"maxContentLevel":27,"version":24},"e36c9c19-4e7d-4c69-a3f8-dcff4f365a62",{"type":27,"summary":3691},[3692,3693,3694],"LAWs are intelligent military systems that can track and potentially kill targets","LAWs are considered a third revolution in warfare, following gunpowder and nuclear weapons","Many AI researchers are calling for a ban on LAWs",{"id":3696,"data":3697,"type":29,"maxContentLevel":27,"version":24},"db79c413-c732-4e68-9446-fe8027ca6ea6",{"type":29,"intro":3698},[3699,3700],"What are LAWs, and what are they used for?","Why are many AI researchers calling for LAWs to be banned?",[3702,3726,3743],{"id":3703,"data":3704,"type":24,"maxContentLevel":27,"version":24,"reviews":3707},"fcc7fb9d-4663-44a6-80aa-31b47b066596",{"type":24,"contentRole":42,"markdownContent":3705,"audioMediaId":3706},"AI is a tool, and in the hands of the wrong people, that tool can be extremely dangerous. Last time, we talked about deepfakes. But another example is the recent rise of **lethal autonomous weapons (LAWs)**.\n\nLAWs are advanced military systems that can be trained to track down human targets, and potentially even kill them. For example, the US military's Replicator Initiative is a cutting-edge plan to build intelligent swarms of unmanned, weaponized drones.\n\n![Graph](image://af3b4a45-14d2-40a5-9482-6124ec6242cc \"Military drone. By АрміяІнформ (CC BY 4.0) \u003Chttps://creativecommons.org/licenses/by/4.0>, via Wikimedia Commons\")\n\nAs things stand, most LAWs still rely on a human to make the final decision to attack. But that could change in the near future. It's why a lot of people refer to LAWs by a different name: 'killer robots'.","068afc66-8be0-4306-bd0b-fd148255255e",[3708,3719],{"id":3709,"data":3710,"type":67,"version":24,"maxContentLevel":27},"89cde1ec-fea8-42c4-b7a9-9c28d612611e",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":3711,"multiChoiceCorrect":3713,"multiChoiceIncorrect":3715,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[3712],"In the context of AI, what are LAWs?",[3714],"Lethal Autonomous Weapons",[3716,3717,3718],"Learning Autonomous Weapons","Learning Artificial Weapons","Lethal Artificial Weapons",{"id":3720,"data":3721,"type":67,"version":24,"maxContentLevel":27},"6f5a5687-555c-4a47-80ba-7f853603977f",{"type":67,"reviewType":37,"spacingBehaviour":24,"clozeQuestion":3722,"clozeWords":3724},[3723],"The US military's Replicator Initiative is a plan to build intelligent swarms of unmanned, weaponized drones.",[3725],"Replicator",{"id":3727,"data":3728,"type":24,"maxContentLevel":27,"version":24,"reviews":3731},"44734f87-7256-4237-a167-324a9461c893",{"type":24,"contentRole":42,"markdownContent":3729,"audioMediaId":3730},"LAWs bring to mind the *I, Robot* stories, written by Isaac Asimov way back in the 40s and 50s. We've talked about Asimov already in this pathway. He's the guy who came up with the Frankenstein Complex – that's the idea that humans are instinctively scared of robots.\n\nAsimov's stories featured fictional robots, which were always programmed with a set of pre-defined rules. He called them the **Laws of Robotics**. They helped to make sure that these fictional robots would always be a positive force in society.\n\nThe First Law of Robotics stated that a robot must never harm a human. In no context were they able to break this rule. It was fundamental to the way these robots behaved.","50fd5f1c-8fae-4ad7-9e9d-6955cd79524e",[3732],{"id":3733,"data":3734,"type":67,"version":24,"maxContentLevel":27},"0f6b27d5-27a6-4862-bc0f-27b02e5fd8de",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":3735,"multiChoiceCorrect":3737,"multiChoiceIncorrect":3739,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[3736],"In Asimov's 'I, Robot' stories, what was the First Law of Robotics?",[3738],"A robot must never harm a human",[3740,3741,3742],"A robot must always obey orders","A robot must protect its own existence","A robot must never lie",{"id":3744,"data":3745,"type":24,"maxContentLevel":27,"version":24,"reviews":3748},"67ae3273-5d4f-4082-afec-b264ac92f68d",{"type":24,"contentRole":42,"markdownContent":3746,"audioMediaId":3747},"In the days of Asimov, the idea of robots harming humans was purely hypothetical. But now it's an essential ethical question that urgently needs to be resolved.\n\nIf you asked most scientists, they wouldn't want AI to be used for weapons. In 2015, several thousand AI researchers signed an open letter calling for a ban on lethal autonomous weapons. But as things stand, these weapons are still being developed.\n\n![Graph](image://0bdc44be-46e0-48b5-a8f3-abb3aeb2ddb6 \"LAW protest. By Campaign to Stop Killer Robots (CC BY 2.0) \u003Chttps://creativecommons.org/licenses/by/2.0>, via Wikimedia Commons\")\n\nThat letter calls it a third revolution in warfare. First, we had the invention of gunpowder. Then we had the invention of nuclear weapons. Now, we have the invention of AI weapons – and warfare might never be the same.","2066ef0d-d670-4c2c-a7c5-3bba9292d7ad",[3749,3755],{"id":3750,"data":3751,"type":67,"version":24,"maxContentLevel":27},"a68441bc-e159-4c27-924c-c53959198d2c",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":3752,"activeRecallAnswers":3754},[3753],"In 2015, several thousand AI researchers signed an open letter calling for a ban on what?",[3714],{"id":3756,"data":3757,"type":67,"version":24,"maxContentLevel":27},"f71c414e-3402-4226-aedf-ec47193228e0",{"type":67,"reviewType":15,"spacingBehaviour":24,"orderAxisType":24,"orderQuestion":3758,"orderItems":3760},[3759],"Some experts think that warfare has gone through three revolutions. Put these revolutions in order:",[3761,3763,3765],{"label":3762,"sortOrder":4},"Gunpowder weapons",{"label":3764,"sortOrder":24},"Nuclear weapons",{"label":3766,"sortOrder":42},"AI weapons",{"id":3768,"data":3769,"type":42,"version":24,"maxContentLevel":27,"summaryPage":3771,"introPage":3778,"pages":3784},"955fba80-735a-43d4-b39e-e115b75afe83",{"type":42,"title":3770},"Energy demands",{"id":3772,"data":3773,"type":27,"maxContentLevel":27,"version":24},"57f86301-2fae-4d80-b37e-b647069807a6",{"type":27,"summary":3774},[3775,3776,3777],"Data centers must be used to host the most powerful AI models","Data centers are resource intensive, and have negative impacts on the planet","AI models can also be beneficial for the planet, but it's unclear whether this outweighs their resource demands",{"id":3779,"data":3780,"type":29,"maxContentLevel":27,"version":24},"1621f527-504b-48ba-bb93-1691156c7e82",{"type":29,"intro":3781},[3782,3783],"What are the energy demands of powerful AI models?","Do the benefits of AI models outweigh these energy demands?",[3785,3800,3817,3839],{"id":3786,"data":3787,"type":24,"maxContentLevel":27,"version":24,"reviews":3790},"cc98be43-412d-4851-ac52-8a6ceeac9834",{"type":24,"contentRole":42,"markdownContent":3788,"audioMediaId":3789},"There's something we haven't really talked about yet. The world's most powerful AI models demand a lot of computing power.\n\nBecause of this, these models can't be hosted by a single, individual computer. Instead, they need to be hosted by something called a **data center**.\n\nA data center is like a giant warehouse full of powerful computers and servers. Working together, these computers can process vast amounts of data. It's the only real way to run a model (and train a model) as powerful and complex as something like ChatGPT.\n\n![Graph](image://8c5625e8-4b92-42e0-8b0c-e1d0fbffd7ed \"Data center. By www.fh-ap.com (CC BY-SA 3.0) \u003Chttps://creativecommons.org/licenses/by-sa/3.0>, via Wikimedia Commons\")\n\nBut here's the thing. These data centers use vast amounts of energy. And that raises another ethical question: how bad is AI for the environment?","adb6fd64-4da5-4132-8381-b74fa99ef91d",[3791],{"id":3792,"data":3793,"type":67,"version":24,"maxContentLevel":27},"b201d741-7f35-43b4-b947-e9093a12f66d",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":3794,"binaryCorrect":3796,"binaryIncorrect":3798},[3795],"What is a data center?",[3797],"A giant warehouse full of powerful computers and servers",[3799],"An institution which collects data for AI companies",{"id":3801,"data":3802,"type":24,"maxContentLevel":27,"version":24,"reviews":3805},"53655bb2-c84e-41e5-abc7-17a13729d713",{"type":24,"contentRole":42,"markdownContent":3803,"audioMediaId":3804},"As things stand, approximately 2% of global energy production is used to power data centers. But by 2030, this number could rise as high as 4%.\n\nThis isn't just down to AI. Data centers are also used to host websites, apps, and so on. But the rapid growth of the AI industry is definitely the driving force.\n\nAccording to a recent estimate, the total energy demands of AI models is doubling every 100 days. By 2030, the AI industry is likely to be using more energy per year than countries like Iceland and the Netherlands.","a852706e-bf42-414d-a0b0-2dd55956ee7d",[3806],{"id":3807,"data":3808,"type":67,"version":24,"maxContentLevel":27},"b6eb2588-e688-4b85-ab93-8ec514550491",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":3809,"multiChoiceCorrect":3811,"multiChoiceIncorrect":3813,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[3810],"According to one estimate, the energy demands of the AI industry are doubling roughly how often?",[3812],"Every 100 days",[3814,3815,3816],"Every 50 days","Every year","Every two years",{"id":3818,"data":3819,"type":24,"maxContentLevel":27,"version":24,"reviews":3822},"eb2ab208-a04c-46ca-827d-27d50c4f9c33",{"type":24,"contentRole":42,"markdownContent":3820,"audioMediaId":3821},"In 2024, Shaolei Ren – a professor of computing in California, USA – actually worked out how much energy it took to run a single ChatGPT query.\n\nSupposedly, if you ask ChatGPT to generate a 100 word email, its data center uses the same amount of power as it would take to power a lightbulb for fourteen hours.\n\nIt's worth pointing out that data centers also use other resources. Water, for example, is used to keep all these powerful computers cool.\n\nAccording to Shaolei Ren, that same 100 word email would also use the equivalent of one bottle of water.","1087e6d7-3fcd-48e8-b751-64937ed50509",[3823,3832],{"id":3824,"data":3825,"type":67,"version":24,"maxContentLevel":27},"c8301c2c-fe27-4ab1-8ac9-8702cd6c98d8",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":3826,"binaryCorrect":3828,"binaryIncorrect":3830},[3827],"A 100-word ChatGPT output uses enough energy to power a lightbulb for how many hours?",[3829],"14",[3831],"1",{"id":3833,"data":3834,"type":67,"version":24,"maxContentLevel":27},"d2b45dec-07f9-4148-b41c-296449bf67e6",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":3835,"binaryCorrect":3837,"binaryIncorrect":3838},[3836],"A 100-word ChatGPT output uses enough water to fill how many bottles?",[3831],[3829],{"id":3840,"data":3841,"type":24,"maxContentLevel":27,"version":24,"reviews":3844},"2115b0c8-9329-46ce-9580-bb3da7487e29",{"type":24,"contentRole":42,"markdownContent":3842,"audioMediaId":3843},"To sum things up: Artificial Intelligence is extremely resource intensive. And that might not be great for the future health of the planet.\n\nOf course, AI models can also be *good* for the planet. Imagine, for example, a model that studies millions of houses, then finds patterns that help us to develop a new type of energy efficient home. Or a model which studies weather patterns, then helps us collect water more effectively.\n\nBut it's hard to know whether these benefits outweigh the resource demands of these models. It's yet another question that needs answering: is Artificial Intelligence bad for the planet or not?\n\n![Graph](image://140eac1f-6454-40a2-bedd-0fbfaa6ae4da \"Our planet. Image: (Public domain), via Wikimedia Commons\")","d24c6669-e2cf-4867-a3c2-b11013bba181",[3845,3856],{"id":3846,"data":3847,"type":67,"version":24,"maxContentLevel":27},"66ffee1f-d755-4955-9359-0673225b22ed",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":3848,"multiChoiceCorrect":3850,"multiChoiceIncorrect":3852,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[3849],"Are AI models good or bad for the planet?",[3851],"Unclear – both of these things are true",[3853,3854,3855],"Good – they can come up with solutions to problems","Bad – they are extremely resource intensive","Unclear – neither of these things are true",{"id":3857,"data":3858,"type":67,"version":24,"maxContentLevel":27},"a2d27ad5-6ead-4a64-8353-41439fcef9d9",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":3859,"activeRecallAnswers":3861},[3860],"By 2030, the AI industry is likely to be using more energy per year than what?",[3862],"Countries like Iceland and the Netherlands",{"id":3864,"data":3865,"type":42,"version":24,"maxContentLevel":27,"summaryPage":3867,"introPage":3874,"pages":3880},"2c23d8c0-908f-4f94-8ea4-aefa7be68c14",{"type":42,"title":3866},"Citizenship & rights",{"id":3868,"data":3869,"type":27,"maxContentLevel":27,"version":24},"9c077543-6aca-4fc3-bea9-6afe06556dbc",{"type":27,"summary":3870},[3871,3872,3873],"Sophia, a humanoid robot, was granted citizenship in Saudi Arabia","This was just a publicity stunt, and ANI models don't really need rights","However, if AGI is ever invented, it will raise many questions about rights",{"id":3875,"data":3876,"type":29,"maxContentLevel":27,"version":24},"42892aae-1d1f-4ad7-af71-9d364dc733d3",{"type":29,"intro":3877},[3878,3879],"Which robot was recently granted citizenship in Saudi Arabia?","If AGI models were ever invented, what ethical questions would that raise?",[3881,3902,3907],{"id":3882,"data":3883,"type":24,"maxContentLevel":27,"version":24,"reviews":3886},"f495677c-76d6-412c-9d23-613cf8a3a239",{"type":24,"contentRole":42,"markdownContent":3884,"audioMediaId":3885},"Here's another ethical question: should AI models have rights?\n\nEarlier, we talked about Sophia – a humanoid robot developed by Hanson Robotics. In 2017, this robot was actually granted citizenship in Saudi Arabia. In theory, this gave her the right to marry and vote.\n\nBut this was more of a publicity stunt than anything. As we mentioned before, Sophia is just an Artificial Narrow Intelligence with a face. Granting her citizenship, and human rights, is essentially no different to granting citizenship to your laptop.\n\n![Graph](image://fe2fbc13-5dbe-4781-8454-5398f7b8abd0 \"Sophia robot. By Satirdan kahraman (CC BY-SA 4.0) \u003Chttps://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons\")","141dcf0a-e015-4a07-b9b2-1baf4e11e088",[3887],{"id":3515,"data":3888,"type":67,"version":24,"maxContentLevel":27},{"type":67,"reviewType":27,"spacingBehaviour":24,"collapsingSiblings":3889,"multiChoiceQuestion":3890,"multiChoiceCorrect":3892,"multiChoiceIncorrect":3894,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6,"matchPairsQuestion":3898,"matchPairsPairs":3899},[3510,3513,3514],[3891],"In a 2017 publicity stunt, which country granted citizenship to the humanoid robot Sophia?",[3893],"Saudi Arabia",[3895,3896,3897],"UAE","United States","China",[179],[3900],{"left":3893,"right":3901,"direction":27},"Granted citizenship to Sophia",{"id":3903,"data":3904,"type":24,"maxContentLevel":27,"version":24},"2cb8b569-cdb6-472d-bfcf-6b981c8b729c",{"type":24,"contentRole":42,"markdownContent":3905,"audioMediaId":3906},"So, today's generation of narrow models don't really need any rights. At least, they don't need them any more than a laptop would need them, or a toaster, or a mobile phone.\n\nThings would get a lot more complicated, though, if we started to develop Artificial General Intelligence (AGI) models, especially if these models showed signs of human-like consciousness and emotions.\n\nHere's a question. As things stand, AI is a tool. But with a human-like mind, does it suddenly become a slave? The word \"robot\" is actually derived from a Czech word, which roughly translates as ‘forced labor’.","80445a03-e592-4f24-a273-354335840140",{"id":3908,"data":3909,"type":24,"maxContentLevel":27,"version":24,"reviews":3912},"14adbfa3-a739-4a27-9fcd-ca3734b3ffbb",{"type":24,"contentRole":42,"markdownContent":3910,"audioMediaId":3911},"As we said at the start of this entire pathway, there's no guarantee that an AGI model will ever be invented. But if it *was* invented, it would open up so many ethical questions that it would be hard to even keep track.\n\nWe've already talked about rights and citizenship. But what about marriage and love? Should people be allowed to have relationships with robots? Should robots have relationships with each other?\n\nThere's also the legal side of things. If you shutdown an AGI, is it murder? And how about religious questions. Does an AGI have a soul?\n\nAgain: none of this is relevant right now. But it's an interesting topic to think about. Narrow AI has already opened plenty of urgent ethical dilemmas, but these are nothing compared to the existential questions that would come with AGI.","0b172acf-96c2-413a-a7b1-3c84416512d7",[3913,3920],{"id":3914,"data":3915,"type":67,"version":24,"maxContentLevel":27},"e1374b23-c955-4ce4-9d7b-589f1bafaf1d",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":3916,"binaryCorrect":3918,"binaryIncorrect":3919},[3917],"Which of these types of Artificial Intelligence would raise the most ethical questions?",[185],[182],{"id":3921,"data":3922,"type":67,"version":24,"maxContentLevel":27},"9afe69e9-59aa-4247-bf19-45108afe2744",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":3923,"multiChoiceCorrect":3925,"multiChoiceIncorrect":3927,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[3924],"The word \"robot\" is derived from a Czech word, which roughly translates as what?",[3926],"Forced labor",[3928,3929,3930],"Free labor","Machine man","Computer man",{"id":3932,"data":3933,"type":28,"maxContentLevel":27,"version":42,"orbs":3936},"e844dbdc-3eae-45ba-967a-f6f9969e772c",{"type":28,"title":3934,"tagline":3935},"The Future","What happens next?",[3937,4028],{"id":3938,"data":3939,"type":42,"version":42,"maxContentLevel":27,"summaryPage":3941,"introPage":3948,"pages":3954},"8e3b5fce-2c8c-4099-bfb8-63a5914e8f42",{"type":42,"title":3940},"The next few years",{"id":3942,"data":3943,"type":27,"maxContentLevel":27,"version":24},"dbc706f4-c875-4b90-91ee-0fbbf4dd4fc6",{"type":27,"summary":3944},[3945,3946,3947],"AI models will probably become more powerful and complex in the future","The development of new robots could be used for a variety of tasks","However, the future of AI is uncertain, due to potential ethical challenges",{"id":3949,"data":3950,"type":29,"maxContentLevel":27,"version":24},"fd53898e-4737-4ee0-a701-82bbc7a76852",{"type":29,"intro":3951},[3952,3953],"What will AI models potentially look like in the future?","Why is there still some uncertainty about the future of AI?",[3955,3960,3973,3988,4011],{"id":3956,"data":3957,"type":24,"maxContentLevel":27,"version":24},"3ab70109-ee24-4aac-acd9-04224f4978c4",{"type":24,"contentRole":42,"markdownContent":3958,"audioMediaId":3959},"At last, we’re reaching the end of this pathway. But it’s important to remember: the story of AI is still only just beginning.\n\nIt’s hard to predict what’s coming next, but over the next few years, there are bound to be lots of developments. Better neural networks, more impressive robots… and with every passing year, more and more people will start using AI as part of their daily lives.\n\nJust think: a few decades ago, barely anyone owned a mobile phone, and even fewer people used the internet. Now, we use these technologies so much that it's hard to believe we ever lived without them.","0bf361b3-b58e-41fc-97eb-3f9126e4c8c7",{"id":3961,"data":3962,"type":24,"maxContentLevel":27,"version":24,"reviews":3965},"20b37712-d38f-4de6-a18c-7d9bcb67d4dc",{"type":24,"contentRole":42,"markdownContent":3963,"audioMediaId":3964},"In 2024, Sam Altman (current CEO of OpenAI) said that the next big step for AI models would be their implementation as multi-purpose personal assistants.\n\n![Graph](image://461a95bd-ede6-4c05-a950-6eae876679be \"Sam Altman. Image: (cropped) by TechCrunch (CC BY 2.0) \u003Chttps://creativecommons.org/licenses/by/2.0>, via Wikimedia Commons\")\n\nImagine an AI on your phone, which Altman described as a \"super-competent colleague that knows absolutely everything about my whole life, every email, every conversation I’ve ever had.\"\n\nThink of all the questions you could ask it: \"Hey, what was the name of the guy I met at that conference last year?\" It could do other things too, like field emails, clean up photos, give advice, and so on.\n\nSomething like this raises more ethical questions (how would you feel about an AI having access to all your personal information?) but it's hard to deny that it wouldn't be an extremely useful tool.","6f24f560-e68c-44d5-a369-ac45af5a9c6f",[3966],{"id":3967,"data":3968,"type":67,"version":24,"maxContentLevel":27},"cecfb46f-f189-449b-95bf-3ff647d8aabe",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":3969,"activeRecallAnswers":3971},[3970],"In 2024, Sam Altman (CEO of OpenAI) predicted that AI would soon be used for what?",[3972],"Multi-purpose personal assistants",{"id":3974,"data":3975,"type":24,"maxContentLevel":27,"version":24,"reviews":3978},"b3324f62-17c4-4a85-9d28-be759cea51b5",{"type":24,"contentRole":42,"markdownContent":3976,"audioMediaId":3977},"In the near future, we're also likely to see more powerful and complex AI models developed for industries like healthcare, law, and education.\n\nAt the moment, doctors waste so much time on note-taking and administration. If these tasks could be done by AI models, it would free the doctors up to spend more time with patients. According to a report by MedTech Europe, this extra time would be the equivalent of hiring 500,000 additional doctors (in Europe).\n\nMeanwhile, the education industry could benefit from a lot more personalization and custom learning for students. AI could keep track of the needs and interests of individual students, and help teachers to provide the lessons that suit them best.\n\nIt could also have a major impact on accessibility. Imagine, for example, an NLP model which could read the whiteboard to students with vision impairment.","cd479b95-9d4c-4105-baf2-139588b740a4",[3979],{"id":3980,"data":3981,"type":67,"version":24,"maxContentLevel":27},"45965eb8-a7b0-4cb3-8677-26578ad31f98",{"type":67,"reviewType":42,"spacingBehaviour":24,"binaryQuestion":3982,"binaryCorrect":3984,"binaryIncorrect":3986},[3983],"AI note-taking could free up so much time in European healthcare, that it would be the equivalent of hiring what?",[3985],"500,000 new doctors",[3987],"100,000 new doctors",{"id":3989,"data":3990,"type":24,"maxContentLevel":27,"version":24,"reviews":3993},"cb76ed3c-ec60-4539-9548-dddf406b18f9",{"type":24,"contentRole":42,"markdownContent":3991,"audioMediaId":3992},"Over the next few years, we're also likely to see some major leaps forward in robotics.\n\nIn particular, we could see the first humanoid robots with genuine real-world uses. Tesla, for example, is currently developing a robot called Optimus, which they describe as a \"general purpose, bi-pedal, autonomous humanoid robot capable of performing unsafe, repetitive or boring tasks.\"\n\n![Graph](image://dd8b1fa8-73e1-4b8f-be5e-92178ea751fd \"Optimus robot. Image: (Public domain), via Wikimedia Commons\")\n\nRobots could also lead to a bit of a revolution in machine learning. According to Yann LeCun (Chief AI Scientist at Meta), there's no bigger dataset than the actual world – and robotics can allow a neural network to experience this data first hand.\n\nTo put it into perspective: the most powerful neural networks in the world right now have been trained on the entire internet. But by the time a real child is four years old, they've experienced roughly 15 times more data. If a robot could have the same experience, its learning would be off the charts.","86aed173-ff93-460f-92a6-39f07d6b807c",[3994,4004],{"id":3995,"data":3996,"type":67,"version":24,"maxContentLevel":27},"a4c89f21-e4eb-48cf-8b63-60070b3a25a1",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":3997,"multiChoiceCorrect":3999,"multiChoiceIncorrect":4001,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[3998],"What is the name of the humanoid robot which Tesla is currently working on?",[4000],"Optimus",[4002,4003,3281],"Robotos","Primus",{"id":4005,"data":4006,"type":67,"version":24,"maxContentLevel":27},"543e3028-5689-4b4f-acf8-950599fb53f2",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":4007,"activeRecallAnswers":4009},[4008],"According to Yann LeCun (Chief AI Scientist at Meta), there's no bigger dataset than what?",[4010],"Exploring the actual world",{"id":4012,"data":4013,"type":24,"maxContentLevel":27,"version":42,"reviews":4016},"2a420fb0-51a4-4ecf-97ad-fa99d393b6b9",{"type":24,"contentRole":42,"markdownContent":4014,"audioMediaId":4015},"A word of caution: there's a lot of hype around AI right now. But there was also a lot of hype about it in the 1960s... then the technology hit a wall, and we plunged into the AI winter.\n\nExperts aren't really expecting that to happen again. But there could be some hurdles in the next few years, especially surrounding those ethical questions we've talked about. For example, if governments crack down on data use or energy consumption, it could slow the industry down.\n\nThere's also public opinion to think about. As we said, there's plenty of hype around AI... but would that change if we felt like neural networks were invading our privacy, or humanoid robots were taking our jobs?\n\nEither way, we have an interesting few years ahead of us.","ceea6aab-8309-40fc-9d02-8a9642e6fe56",[4017],{"id":4018,"data":4019,"type":67,"version":42,"maxContentLevel":27},"235aa18f-f93b-4fc2-a356-e6486fbfb807",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":4020,"multiChoiceCorrect":4022,"multiChoiceIncorrect":4024,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[4021],"By the time a child is four years old, they've experienced approximately how much data compared to the world's most powerful neural networks?",[4023],"The child has experienced 15 times more data",[4025,4026,4027],"The child has experienced 5 times more data"," The child has experienced 10 times more data"," The child has experienced 20 times more data",{"id":4029,"data":4030,"type":42,"version":24,"maxContentLevel":27,"summaryPage":4032,"introPage":4040,"pages":4046},"115425d1-3b4f-40eb-a496-364e7d9c30be",{"type":42,"title":4031},"Singularity",{"id":4033,"data":4034,"type":27,"maxContentLevel":27,"version":24},"0a76fff1-1ffe-4cdb-83bf-145795fe2034",{"type":27,"summary":4035},[4036,4037,4038,4039],"The singularity is a potential moment when AI surpasses humans","Beyond this point, AI models could recursively self-improve","Eventually, they'll develop Artificial Super Intelligences","If these ASIs saw humans as competition, they could potentially destroy our species",{"id":4041,"data":4042,"type":29,"maxContentLevel":27,"version":24},"1245ad18-0edf-4aca-a61d-d9f490544a17",{"type":29,"intro":4043},[4044,4045],"What is the singularity, and why could it threaten humanity?","Do we really need to be worried about this, or not?",[4047,4064,4077,4101],{"id":4048,"data":4049,"type":24,"maxContentLevel":27,"version":24,"reviews":4052},"585c2d7c-bb81-45c9-a2e4-2d7bbb9c4e97",{"type":24,"contentRole":42,"markdownContent":4050,"audioMediaId":4051},"While we're talking about the future of AI... there's one more thing we should mention. Experts call it the **singularity** – and some of them think it's the greatest threat to life on Earth as we know it.\n\nSpecifically, this refers to a moment in time when an AI model reaches a level of intelligence which is greater than the intelligence of humans. Think of human intelligence, and Artificial Intelligence, as two lines on a graph. The singularity is the point where the two lines cross.\n\n![Graph](image://c0fa6866-9e70-43b3-a729-4941456fa5f1 \"The singularity.\")\n\nAs we've mentioned repeatedly in this pathway so far, that moment is surely still a long way off, and there's no guarantee that it will ever even happen at all. But let's assume, for a moment, that it *will* happen. Why would that be a threat to life on Earth?","7a1e2c26-e626-4002-8a68-e11778ee3c23",[4053],{"id":4054,"data":4055,"type":67,"version":24,"maxContentLevel":27},"094dfbb2-892f-4abe-9c66-26ff34f26fef",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":4056,"multiChoiceCorrect":4058,"multiChoiceIncorrect":4060,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[4057],"What do we call the theoretical moment when the first AI model surpasses human intelligence?",[4059],"The singularity",[4061,4062,4063],"The transcendence","The break","The anomaly",{"id":4065,"data":4066,"type":24,"maxContentLevel":27,"version":24,"reviews":4069},"4b992ec1-c2eb-4877-a5d7-273dc3316e40",{"type":24,"contentRole":42,"markdownContent":4067,"audioMediaId":4068},"Here's the thing. If humans created an AI model that was slightly more intelligent than us... then that AI would surely be smart enough to create an AI of its own.\n\nThat second AI would be even smarter. After all, it was made by a slightly better creator. And in turn, that second, smarter AI could create an even smarter model. This process would repeat again and again, with each AI being smarter than the last.\n\nScientists call this **recursive self-improvement**. In a matter of years, we could end up with Artificial Super Intelligences which are thousands of times smarter than humans.","8647ef51-8cca-4402-9d0e-d09f5497fc2f",[4070],{"id":4071,"data":4072,"type":67,"version":24,"maxContentLevel":27},"5d9cdc0a-6281-47a4-a2c6-72fb1d880364",{"type":67,"reviewType":24,"spacingBehaviour":24,"activeRecallQuestion":4073,"activeRecallAnswers":4075},[4074],"What do we call a theoretical chain of Artificial Intelligences which develop smarter and smarter models?",[4076],"Recursive self-improvement",{"id":4078,"data":4079,"type":24,"maxContentLevel":27,"version":24,"reviews":4082},"2e07bb65-4416-438a-89bb-04e992c74161",{"type":24,"contentRole":42,"markdownContent":4080,"audioMediaId":4081},"Now, we have to assume that these self-improving Artificial Super Intelligences would require a lot of resources. How else would they power their systems?\n\nAnd that's why all this could turn out to be a threat to life on Earth. If these models saw humanity as competition, there's a chance they would wipe us out. Why would they bother keeping us around?\n\nIn 2014, Professor Stephen Hawking said this: \"Humans, who are limited by slow biological evolution, couldn't compete, and would be superseded.\" Elon Musk, meanwhile, has called this phenomenon \"our biggest existential threat.\"","5b405db2-21fd-4fbb-a6ae-4d672f305eab",[4083,4090],{"id":4084,"data":4085,"type":67,"version":24,"maxContentLevel":27},"c9dba846-1b98-4114-9c26-243f49ba0812",{"type":67,"reviewType":37,"spacingBehaviour":24,"clozeQuestion":4086,"clozeWords":4088},[4087],"In 2014, Stephen Hawking said that the singularity would eventually lead to humans being \"superseded\".",[4089],"superseded",{"id":4091,"data":4092,"type":67,"version":24,"maxContentLevel":27},"637f8760-43b3-4329-8b39-8d7fad2067ee",{"type":67,"reviewType":27,"spacingBehaviour":24,"multiChoiceQuestion":4093,"multiChoiceCorrect":4095,"multiChoiceIncorrect":4097,"multiChoiceMultiSelect":6,"multiChoiceRevealAnswerOption":6},[4094],"How did Elon Musk describe the idea of an AI singularity?",[4096],"\"Our biggest existential threat.\"",[4098,4099,4100],"\"Our greatest achievement.\"","\"Our only hope of survival.\"","\"Our final chapter.\"",{"id":4102,"data":4103,"type":24,"maxContentLevel":27,"version":24},"48a6ff95-b27d-4240-b4d5-c23491afe99c",{"type":24,"contentRole":42,"markdownContent":4104,"audioMediaId":4105},"Just to repeat this one more time: there's no guarantee that we'll ever reach the singularity. Even if we do, the idea of recursive self-improvement eventually leading to human extinction is totally hypothetical.\n\nBut it's another important ethical question that we need to bear in mind: if we start to get close to the singularity, is it a good idea to continue?\n\nFor now, there are so many reasons to be excited about the rise of AI. That's what you should try to focus on as you come to the end of this pathway. But at the same time, it's important to reserve a little bit of caution.\n\nIt's the same with any new technology. While enjoying the benefits, we also need to prepare ourselves for challenges along the way.\n\n![Graph](image://d789f080-7aaa-4973-ab8a-2b877bc72dc4 \"Robot. Image via Pexels\")","bfcdbdf1-f808-421a-8b71-f92cd3769754",{"left":4,"top":4,"width":4107,"height":4107,"rotate":4,"vFlip":6,"hFlip":6,"body":4108},24,"\u003Cpath fill=\"none\" stroke=\"currentColor\" stroke-linecap=\"round\" stroke-linejoin=\"round\" stroke-width=\"2\" d=\"m9 18l6-6l-6-6\"/>",{"left":4,"top":4,"width":4107,"height":4107,"rotate":4,"vFlip":6,"hFlip":6,"body":4110},"\u003Cg fill=\"none\" stroke=\"currentColor\" stroke-linecap=\"round\" stroke-linejoin=\"round\" stroke-width=\"2\">\u003Cpath d=\"M12.586 2.586A2 2 0 0 0 11.172 2H4a2 2 0 0 0-2 2v7.172a2 2 0 0 0 .586 1.414l8.704 8.704a2.426 2.426 0 0 0 3.42 0l6.58-6.58a2.426 2.426 0 0 0 0-3.42z\"/>\u003Ccircle cx=\"7.5\" cy=\"7.5\" r=\".5\" fill=\"currentColor\"/>\u003C/g>",1778179171267]