Superintelligence

A hypothetical agent that possesses intelligence far surpassing that of the brightest and most gifted human minds.

How it works

Superintelligence is a theoretical concept rather than a current technology. It describes an AI system that can outperform all humans in every cognitive domain — scientific reasoning, social intelligence, creativity, and strategic planning. Paths hypothesised to lead to superintelligence include recursive self-improvement (an AI that can iteratively improve its own design), whole brain emulation, or simply scaling up current architectures far beyond current sizes. Its capabilities and timeline are deeply uncertain.

Why it matters

Superintelligence is the central concern of the AI safety movement. If a system could outthink humanity in every domain, it would gain decisive strategic advantage, potentially making it impossible for humans to correct any misaligned goals it has. This concern motivates research into alignment, interpretability, and governance frameworks. Whether or not superintelligence is achievable or imminent, the possibility shapes the policy priorities of major AI labs and governments worldwide.