AI glossary
AGI (artificial general intelligence)
A hypothetical AI able to match or exceed humans at most cognitive tasks, not just specific ones.
Today’s systems are impressive but uneven: superhuman at some tasks, unreliable at others. AGI describes the point where an AI could do most intellectual work a person can, and learn new tasks as flexibly.
There is no agreed definition or test, so claims about when AGI will arrive vary widely and often say as much about the speaker as about the technology. Treat precise predictions with caution.
Example: A current model can solve a maths olympiad problem and, in the same conversation, make a mistake on a simple task that a person would not. AGI, as it is usually defined, would not have such uneven gaps.
In practice
- When you read that AGI will arrive by a certain date, check who says so and how they define it.
- To decide whether to use AI in your work, what models do well today matters more than predictions.
- “Superintelligence” describes an AI that would outperform people at almost everything; it has no single definition either.
To understand what models can do today, start with how LLMs work.