Against p(doom): assigning probabilities to one-shot events is sillier than it sounds
GabGarrett · x · 2026-11-11
The author, self-describing as having gotten a D in probability and statistics, explains why AI-apocalypse p(doom) discourse never swayed him: assigning probabilities to one-shot events is inherently grotesque. Card draws and equipment failures allow probability definitions; humanity-level single-shot outcomes don't.
His examples: the atomic bomb team was told there was a probability of igniting Earth's atmosphere and built it anyway; moon-landing and Cold War nuclear-war forecasts similarly never stopped action. The post supports a quoted critique of EAs and Bay Area rationalists who use "p(doom)", "priors" and "bayesians" in pseudo-mathematical ways.
More from AGI Musings
- François Fleuret to biologists: look how cocksure other fields were before AI — francoisfleuret · 2026-11-11
- Could Recursive AI Improvement Compress Millennia of Progress into One Decade? — VraserX · 2026-11-11
- AI 2027 authors say reality is running 5 months ahead of their scenario — LeviTurk · 2026-11-11
- teortaxesTex: RL may give coding models 'full fluency' with no superhuman headroom — teortaxesTex · 2026-11-11
- Frontier AI Still Lacks 'Heresy': Breakthroughs Come from Defying Consensus — JeremyNguyenPhD · 2026-11-11
- Dev argues optimism is the only trait that matters as intelligence gets commoditized — zack_overflow · 2026-11-11