Strogatz to Wolfram: your Prisoner's Dilemma tournament stopped before evolution kicked in
stevenstrogatz · x · 2026-10-06
Mathematician Steven Strogatz published a careful open response to Nassim Taleb analyzing Stephen Wolfram's recent finite-state-machine Prisoner's Dilemma experiments, which found complicated strategies vastly outperforming Tit for Tat.
His core argument:
- He doesn't dispute the computation, nor defend game theory — he only compares simulations to simulations
- Wolfram enumerates all finite-state strategies of a given size and ranks them by performance against the full uniform field
- The key flaw: there is no next generation. Successful strategies don't become more abundant, so it amounts to one round of an evolutionary tournament
- The original population is full of 'suckers' (unconditional cooperators), so strategies skilled at exploiting them score spectacularly in round one
- Once scores drive a new generation and the tournament repeats, round-1 winners need not win round 10 or 1,000 — strategies create the environment in which later selection happens
He invokes Nowak & Sigmund's decades-old methodological work as precedent. The ideas are Strogatz's; the prose was polished by ChatGPT.
More from AGI Musings
- NYMag Deep-Dive on Anthropic Resignation: How Lab Race Cultures Converge on the Same Unsafe Dynamics — Miles_Brundage · 2026-10-06
- Researchers Push Back on Scott Aaronson: Alignment Is Solvable RL, Not Theology — sebkrier · 2026-10-06
- Long Read: Sex, AI, and the Apocalypse Traces the Fringe Roots of AI Doomers — ZeroStateReflex · 2026-10-06
- Built in Three Prompts and an Hour: Why "I Wanted This to Exist" Now Justifies Software — TheMoonMidas · 2026-10-06
- Runway CEO invokes Jevons paradox: coding agents are expanding software demand, video is next — c_valenzuelab · 2026-10-06
- AI researcher: we may not need to wait until next summer for RSI — coherence · 2026-10-06