Artificial general intelligence, or AGI, is one of the most discussed and least consistently defined terms in the AI field. Understanding what researchers actually mean by it — and where they disagree — makes the term far more useful than treating it as a single, agreed-upon milestone.
The core idea
Most definitions of AGI center on an AI system that can perform any intellectual task a human can, across arbitrary domains, rather than excelling only at narrow, specifically-trained tasks. Today’s AI systems, however capable, are still generally specialized — a model exceptional at coding isn’t automatically exceptional at, say, physical reasoning or long-term planning in a novel environment.
Why experts disagree about it
There’s genuine disagreement in the field about what would actually count as achieving AGI, how it should be measured, and how close current systems are to it — estimates from credible researchers range widely, from a few years to several decades, reflecting real uncertainty about the field’s trajectory rather than a settled consensus being simplified for headlines.
Why the term matters despite the disagreement
Even without a settled definition, AGI functions as a useful reference point for discussing the trajectory of AI capability — where current systems fall short of general human-level performance, and what kinds of breakthroughs might be needed to close that gap. Understanding it as an open research question, rather than an imminent, well-defined event, is a more accurate way to follow the ongoing debate.