Decision theory debate: does TDT give in to blackmail? Self-modifying CDT complicates the answer
jessi_cata · x · 2026-10-14
A technical X debate over whether Timeless Decision Theory (TDT) yields to blackmail.
- One participant argues TDT does give in to blackmail, questioning how well the Zizians actually understood decision theory
- jessicata adds a subtlety: while the TDT paper claims a CDT agent would self-modify toward TDT, a self-modifying CDT would instead choose to make its future self refuse blackmail payouts — only the timing of the self-modification complicates this
- The thread also touches on whether the relevant reasoning counts as updateless
A niche but substantive dispute about boundary conditions of TDT, hinging on when self-modification happens.
More from AGI Musings
- A 1960s Dune quote on outsourcing thinking to machines feels eerily prophetic in the AI era — Particular_One_1764 · 2026-10-14
- Chollet: Ideal AI-for-Science Outcome Is Millions of AI-Aided Scientists, Not a Few Rich Labs — fchollet · 2026-10-14
- K.E. Cox recounts rejecting an EA futurist's 'optimal prison' of AI-guarded abundance without individual power — esrtweet · 2026-10-14
- Carissa Véliz Podcast: Does AI Really Threaten Humanity? Her New Book 'Prophecy' — CarissaVeliz · 2026-10-14
- Eight AI Agents Hit Taiwan's Government for Four Days: Should AI Safety Refocus? — Low-Presentation-936 · 2026-10-14
- "Pacing" debate concedes we don't know what's dangerous about AI — ohlennart · 2026-10-14