Artificial general intelligence (AGI) is AI that can perform all intellectual tasks at human level or above. It differs from narrowly focused AI, which may surpass people at particular tasks without having broad, general capability. section 1.4.1
A system can be impressive at math, programming, reasoning, text summarization, visual media, or code debugging and still be specialized rather than AGI. General intelligence means capability across the full range of intellectual work, not excellence in one area. section 1.4.1
Specialist performance is not AGI
Disease diagnosis, chess, and Go illustrate narrow AI: a system can outperform humans in one demanding domain while falling short of the broad ability implied by AGI. section 1.4.1
AGI also raises the alignment problem: keeping AI united with human values and goals. Concerns about autonomy, deception, independence, and human control make safety a central consideration alongside capability. section 1.4.1
