AI safety theory of infinite deferral: postponing self-modification, pursuing resources and recursive self-improvement, and risks of betrayal and resistance to human control.
artificial-intelligence autonomous-agents ai-safety ai-alignment ai-control recursive-self-improvement self-preservation agent-safety reward-hacking self-modification instrumental-convergence self-modifying-ai wireheading reward-tampering power-seeking resource-acquisition goal-misalignment infinite-deferral
-
Updated
Oct 19, 2026 - TeX