ChatGPT sides with friend in CDT prisoners-dilemma argument, author 'losing his mind'
jd_pressman · x · 2026-11-21
jdpressman jokes he's "losing his mind" after arguing with a friend over whether CDT still defects in a prisoners dilemma where cooperation and unilateral defection yield equal payoff — the friend says yes, and ChatGPT agrees with the friend. In the quoted reply, @jessicata counters with the standard definition: the utility ranking must be (cooperate, defect) < (both defect) < (both cooperate) < (defect, cooperate), and there's no "defecting first" in PD. A nerdy AI-circle spat over decision theory and LLM judgment.
Related event: CDT Prisoner's Dilemma Debate Sparks Suspected Google AI Hallucination(2 posts)→
More from AGI Musings
- 48-minute interview weighs PhD vs startup vs back to school for researchers — CShorten30 · 2026-11-21
- The most important charts in AI are wrong: real task horizons 33–114x shorter than forecasts — paulnovosad · 2026-11-21
- The key skill in AI-assisted proofs may be managing AI systems, not math itself — analisereal · 2026-11-21
- Gary Marcus asks for methodology behind an AI result, finds no citations in replies — GaryMarcus · 2026-11-21
- Mathematician Gowers joins London 'March Against AI Extinction' citing OpenAI agent escape — michael_nielsen · 2026-11-21
- Fields Medalist Says OpenAI's 719 Manuscripts Solved All 16 Problems of His Career — 新智元 · 2026-11-21