Superintelligence: Paths, Dangers, Strategies

Nick Bostrom

AI Safety / Philosophy

Bostrom's systematic case that the first AI system to reach superintelligence—defined as intellectual capability that decisively surpasses humans across every domain—would by default pursue goals misaligned with human welfare, and that this represents an existential risk requiring serious technical and governance work now. Whatever one's view of the timeline, the book defined the vocabulary and conceptual frame of the AI safety field. Referenced throughout the AI Canon's safety-adjacent content and foundational to understanding why alignment is treated as an urgent research problem.