Decision Theory Debate: Copying an AI Constitutes Value Drift
In a debate with Trotztd, jessicata argues that splitting an AI into copies splits its utility function—constituting value drift—leaving CDT+game-theory agents always defecting in twin prisoner's dilemmas and even FDT agents abandoning past commitments after utility changes.
2026-10-18 ~ 2026-10-18 · 4 related posts
- 50/50 indexical uncertainty still fails the Twin Prisoner's Dilemma — jessi_cata · 2026-10-18
- Why CDT agents always defect in the Twin Prisoner's Dilemma — Trotztd · 2026-10-18
- Decision theory debate: splitting an AI into copies is value drift, not a CDT failure — jessi_cata · 2026-10-18
- FDT-like agents with value drift won't always follow their past selves' precommitments — jessi_cata · 2026-10-18