Artificial General Intelligence

Artificial general intelligence (AGI) refers to hypothetical AI systems that match or exceed human cognitive abilities across the full range of tasks—reasoning, learning, creativity, and adaptation—rather than excelling at narrow domains like image recognition or game-playing. Current AI is narrow: powerful within specific domains but lacking the generalization, common sense, and flexibility of human cognition. AGI remains speculative; experiments explore scaling, architecture, and training regimes that might lead toward more general capabilities. Applications, if realized, could include scientific discovery, governance, creative work, and existential risk or benefit depending on alignment and control.

The path to AGI is unclear: some argue scaling current architectures will suffice; others believe fundamental breakthroughs are required. Research spans scaling laws, multi-task and meta-learning, embodied cognition, and hybrid symbolic-neural systems. Safety and alignment—ensuring AGI benefits humanity—are central concerns. AGI remains a long-term research horizon; timelines range from decades to never, depending on assumptions.