15:37
2026-06-21
lesswrong.com
ai-safety
A high-level model of AI bargaining
Advanced AIs may use credible commitments unavailable to humans when bargaining over resources, according to a new model based on program equilibrium. The model outlines a two-phase process where agen…