Artificial General Intelligence

Artificial General Intelligence, or AGI, refers to an AI system that can learn and reason across a wide range of tasks at roughly human level, rather than performing well only in the narrow domain it was built for. The contrast is with narrow AI, which describes essentially every system deployed today: a model may write code, diagnose images, or translate fluently, yet each capability comes from training aimed at that kind of work. AGI implies transferring knowledge to genuinely unfamiliar problems, setting sub-goals, and learning from limited experience the way a person does. It has not been achieved, and there is no agreed test for declaring it. OpenAI states AGI as its founding mission, DeepMind and Anthropic frame their work around similar long-horizon goals, and researchers disagree publicly about whether scaling current language models leads there or whether missing ingredients remain. The pitfall is definitional drift: the term gets applied to any impressive new model, which makes claims unfalsifiable and muddies policy debate. AGI is also distinct from ASI, or superintelligence, which describes capability beyond human level, and much of the AI-safety field is concerned with the transition between the two.