Artificial General Intelligence

"Artificial general intelligence" (AGI) is a contested term for AI systems with broad, human-like cognitive capabilities rather than narrow task-specific performance. The phrase was coined in the early 2000s to revive the original AI ambition of creating intelligence as a unified, domain-general phenomenon .

Definitions diverge sharply. DeepMind co-founder Demis Hassabis frames AGI as systems that "should be able to complete almost all cognitive tasks that humans can do," while OpenAI describes it as "highly autonomous systems that outperform humans at most economically valuable work"—notably excluding physical tasks . A 5Y Capital translation of a *Science* piece by Melanie Mitchell highlights how AI researchers and cognitive scientists fundamentally disagree on the nature of intelligence itself, with some arguing biological intelligence is inseparable from embodied experience and therefore unachievable in machines . A commenter, Xu Bowen, pushes back: most AGI researchers agree AGI is not equivalent to superintelligence, and emphasize systems with multiple, potentially conflicting goals learned from environment rather than single-objective optimization .

In practice, the term has become a strategic lodestar. OpenAI and DeepMind both cite AGI as core mission ; at the 2026 WEF, Hassabis and Anthropic's Dario Amodei discussed institutional readiness for "the day after AGI" rather than technical roadmaps . The concept also shapes competitive positioning—DeepSeek, for instance, has been urged to explore architectures "beyond Transformer" and "beyond autoregressive models" as paths toward AGI .