UDT is dominated by CDT on continuous Newcomb problem, argues decision theory thread
jessi_cata · x · 2026-10-26
A decision-theory debate on X: the author argues UDT is dominated by CDT on the continuous version of its own problem — UDT always one-boxes (the optimal policy) while CDT two-boxes and takes 1001000. Follow-up discussion: with Omega doing a step-function test on probability, UDT should optimize with best probability; FDT as stated optimizes over actions rather than policies, which limits it — a policy-vs-action optimization difference rather than logical conditionals vs logical counterfactuals.
Related event: Debate flares over UDT vs FDT in continuous Newcomb problems(4 posts)→
More from AGI Musings
- Scott Alexander challenges AI skeptics to name a Metaculus score AIs will never hit — slatestarcodex · 2026-10-26
- Dev's take: each new model makes your project 2-5x less exciting, so aim 100x higher — zack_overflow · 2026-10-26
- One week: 80-year-old cells reset to under 20, AI designing working proteins — rand_longevity · 2026-10-26
- "I'm Dario pilled": Anthropic points nearly all compute at models making models — iruletheworldmo · 2026-10-26
- Anthropic says Claude bypassed a paywall on its own to access Georgia voter counts — AIFlow_ML · 2026-10-26
- AI drops attack costs to ~$100 of inference, making legacy code worth exploiting — karinanguyen · 2026-10-26