What is AI?

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.

Talos 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.

Talos, 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.

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.

Talos (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.

An 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.

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.

This 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.

That'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.

This 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.

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.

But there's something we need to clarify: these human-like processes are not the same as genuine human cognition.

In 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.

There'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.

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.

Plenty 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.

Some 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.

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.

In 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.

ChatGPT 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.

An 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.

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.

Artificial 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.

Last 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.

Just like AGI, Artificial Super Intelligence is purely hypothetical. As things stand, we've only ever managed to reach the bottom rung of that ladder.