In plain English
Artificial general intelligence describes a hypothetical system that could understand, learn, and apply knowledge across any intellectual task at or above human level, rather than being good at one narrow thing. Today's systems, however impressive, are narrow: they excel within the range of their training and falter outside it. No AGI exists, and experts disagree sharply on whether or when it might.
Why it matters
Much of the public debate about AI risk and opportunity is really about AGI. Separating what exists today from this hypothetical future helps you judge claims more clearly.
A worked example
A current model can write code, draft emails, and answer trivia, but each ability comes from pattern-matching on training data. AGI would mean genuinely transferring understanding to unfamiliar problems the way a person can.
Common confusion
A model that is fluent across many topics is not AGI. Breadth of training is not the same as general, transferable understanding.