AGI Ruin: A List of Lethalities

Published in 2022, 'AGI Ruin: A List of Lethalities' is Eliezer Yudkowsky's detailed technical and philosophical argument that artificial general intelligence poses an existential risk to humanity. The work systematically catalogs dozens of specific failure modes, alignment problems, and scenarios where AGI development could lead to civilizational collapse or extinction.

Yudkowsky argues that current approaches to AI safety are inadequate and that the default trajectory of AGI development leads to catastrophic outcomes. The essay covers technical challenges in value alignment, the difficulty of specifying human values, instrumental convergence, and the speed at which superintelligent systems might act.

The document became a focal point in AI safety debates, influencing discussions at major AI labs and among policymakers. It represents a maximalist position on AGI risk within the effective altruism and AI safety communities, arguing that the problem is harder and more urgent than mainstream AI research acknowledges.