A theoretical form of AI with human-level cognitive flexibility — capable of understanding, learning, and solving any intellectual task that a human can, across any domain.
In Depth
Artificial General Intelligence describes a hypothetical AI system that matches or surpasses human cognitive flexibility. Unlike today's Narrow AI systems — each optimized for one task — an AGI would transfer knowledge across domains, learn from minimal examples, reason abstractly, and adapt to novel situations just as a human can. It is a system that does not merely perform tasks but understands them.
There is no consensus on how to achieve AGI, when it might arrive, or even precisely what it means. Some researchers argue that scaling current Large Language Models with enough data and compute will produce AGI-like capabilities. Others contend that fundamentally new architectures — perhaps inspired by neuroscience or symbolic reasoning — are required. A minority believe AGI is impossible in principle with silicon-based computation.
The stakes around AGI are enormous. Proponents argue it could solve humanity's most pressing problems — disease, climate change, poverty — by compressing decades of scientific progress into years. Critics warn that an AGI whose goals are even slightly misaligned with human values could pose existential risks. This tension drives the field of AI Alignment and AI Safety, both of which consider AGI their primary concern.
AGI remains theoretical — no system today comes close to true human-level generalization. Its eventual arrival, if it happens, would represent the most transformative technological event in human history.