50/50 indexical uncertainty still fails the Twin Prisoner's Dilemma
jessi_cata · x · 2026-10-18
@jessicata responds in the thread: even if an agent facing a Newcomb-style problem has a 50/50 chance of being the original or the copy — so it "should one-box," matching what it would have committed to ex ante — this reasoning still fails the Twin Prisoner's Dilemma, since each copy believes it cannot unilaterally change the outcome and defects anyway. The quoted tweet adds context: she studies imperfect-recall games and the Sleeping Beauty problem because they sketch connections across anthropics, decision theory, reinforcement learning, and consciousness, though a full formal story remains unfinished.
Related event: Decision Theory Debate: Copying an AI Constitutes Value Drift(4 posts)→
More from AGI Musings
- Nick Bostrom on Roman Forum Podcast: "AI Is Already Conscious" — romanyam · 2026-10-18
- Prediction: Replacing a $50K Worker With AI Costs $100K and Delivers Negative ROI — sidjustice_ · 2026-10-18
- 'That is how early we are': most users still don't know agentic AI like Claude Cowork exists — dnlkwk · 2026-10-18
- Miles Brundage: No AI company is secure against state attackers — Miles_Brundage · 2026-10-18
- Fields Medalist Timothy Gowers: one-shotting a proof with ChatGPT means we don't understand it — haider1 · 2026-10-18
- When users write their own apps, the ad economy gives way to token-based economics — xiaosun86 · 2026-10-18