Mental BoundAGI is the idea of an AI that could handle any intellectual task a human can — learn a new language, switch from writing code to diagnosing illness to composing music — and adapt to novel situations without being retrained. Today's AI is narrow by contrast: strong at specific tasks, not general across domains. AGI does not exist, and researchers disagree on whether it is decades away, centuries away, or achievable at all.
The concept matters because it concentrates the field's safety questions. A narrow chatbot that gets facts wrong is an annoyance; a generally capable system that misbehaves would be far more consequential. Much of AI safety research focuses on keeping systems aligned with human values as they become more capable.
For now, AGI is a concept, not a product — treat claims of "AGI" or "human-level AI" with skepticism. Current systems are powerful tools, but narrow ones.