AGI (artificial general intelligence): what it is and whether it exists yet

Artificial general intelligence (AGI) is a still-hypothetical AI that could learn and do any intellectual work at a human level or better, not just individual tasks.

Narrow AI versus general AGI

In plain words

Today's AI is like specialist athletes: one is a great swimmer, another a brilliant chess player. AGI would be an all-rounder who masters any new sport in a week, from football to figure skating, without years of training for each. No such "athlete" exists yet, but the biggest labs are working on it.

How it works

Modern models can already do many different things, but they fail where a person would cope easily.
AGI would flexibly transfer knowledge from one field to another.
It would learn new things quickly from few examples and plan far ahead.
There is no single agreed definition or test that would prove AGI has arrived.
That is why estimates of when it will appear vary widely - from "soon" to "not for a very long time".

The road to AGI: where we are now and why the end is in fog

Where you'll see it

You will hear about AGI in the news and in statements from AI companies, as well as in debates about AI safety and its impact on jobs and society. You will not find the word in device specs.

Does it matter when buying

It does not affect what you buy. But it helps you understand the news: when someone claims "we have almost built AGI", ask how exactly they define it. And whenever AGI arrives, it already matters today to learn to use AI wisely and critically.

Myths

Myth: Today's chatbots are already AGI. Reality: Most researchers think not: they are strong at many tasks but have clear weak spots.
Myth: AGI will necessarily be conscious. Reality: Intelligence and consciousness are different things. AGI is defined by the ability to do tasks.

FAQ

How is AGI different from ordinary AI?

Ordinary AI is narrow; AGI would be general, at a human level.

What is "superintelligence" (ASI)?

A hypothetical AI that far surpasses humans in most fields.

Is AGI dangerous?

This is seriously debated, which is why many labs and governments are working on AI safety rules.