Fundamentals

AGI (Artificial General Intelligence)

Hypothetical AI with the ability to understand or learn any intellectual task that a human being can perform.

Definition

Artificial General Intelligence refers to AI systems with the broad, flexible cognitive abilities of humans: learning any new task from minimal examples, transferring knowledge across domains, reasoning about novel problems, and setting and pursuing complex goals. All deployed AI today is "narrow" — excelling in specific domains but failing catastrophically outside them. AGI has neither been achieved nor formally defined.

Definitions and timelines vary enormously among researchers. Some (Demis Hassabis, Sam Altman) suggest AGI may arrive within years; others (Gary Marcus, Yann LeCun) argue current architectures are fundamentally insufficient and the path is much longer.

AGI's potential arrival is a central concern of AI safety research. Transformatively capable AI systems would likely be enormously economically valuable but could also concentrate power, cause labour market disruption, or — if misaligned — pose existential risks.

Examples

  • No verified examples exist
  • OpenAI's stated mission is developing beneficial AGI
  • DeepMind's Gemini was claimed by some as approaching AGI on benchmarks