Artificial general intelligence (AGI) usually refers to AI that can match or exceed human ability across a wide range of cognitive tasks, rather than excelling only in narrow domains.
No Single Definition
Definitions vary:
- AI that can perform most economically valuable work as well as people.
- AI that can learn any intellectual task a human can.
- AI with flexible reasoning, learning and adaptation in unfamiliar situations.
These lead to different judgements about how close we are.
Narrow Versus General
Traditional AI systems were narrow: a chess engine couldn't recognise faces. Today's large models are far more general — they write, code, analyse images and reason across topics — yet still show gaps: unreliable reasoning in places, difficulty learning continuously, and errors people wouldn't make.
Why Definitions Matter
Claims that AGI is near, far or already here often rest on different definitions. When reading such claims, ask what's being measured.
A Spectrum, Not a Switch
Many researchers now think of generality as a matter of degree, measured across many capabilities, rather than a single threshold.
Why It Matters
Highly general AI could transform science, the economy and society, bringing major benefits and serious risks — which is why it attracts so much attention.