AI glossary · Beginner · Safety & alignment
AGI
AGI (artificial general intelligence) refers to AI that can do most of the intellectual work people do, not just one specific task, as well as or better than humans. There is still no agreed-upon precise definition or test for it.
About this term
- Level
- Beginner · Core ideas for first-time readers
- Category
- Basics
- Fields
- Safety & alignment
- Added
- 2026-10-08
What it means
Most AI so far has been "narrow AI" that is good at a defined task, such as translation, image recognition or Go. AGI goes further and refers to AI that can learn and apply knowledge to solve unfamiliar problems at a human level across many fields. The key is not skill in one field but "generality," that is, the ability to generalize to new situations. A level far beyond humans is sometimes called "superintelligence (ASI)."
Why it matters
As LLMs and reasoning models have advanced quickly in writing, coding, math and other fields, AGI has gone from distant fantasy to a goal that AI companies state publicly. Forecasts of when AGI will arrive strongly influence discussions about investment, jobs and policy. That is why alignment and AI safety research are often discussed together with the question of how to control a system as powerful as AGI once it appears.
Things to know
Because AGI has no agreed definition, companies and researchers use different criteria. Some define it as doing most economically valuable work; others as the ability to learn and adapt on its own like a person. So any claim that "AGI has been reached" should be read together with the definition used. A high benchmark score does not by itself mean AGI, and current models still show limitations such as hallucinations and mistakes in unfamiliar situations.
Example
AGI is both a research goal and a name used for corporate teams. As in a news item on this site reporting that Boston Dynamics hired as its new CEO the person who led Amazon's Alexa and AGI organization, large tech companies have dedicated AGI-branded teams and are investing in developing more general-purpose models.