Why p(doom) is a flawed idea: unique events have no predictive probabilities
banteg · x · 2026-09-18
banteg amplified a book excerpt arguing p(doom) is conceptually flawed: probability assignments imply statistical characteristics drawn from accumulated past data, but an AI doom scenario is a unique event with no such data — so the number has no predictive value and merely creates an "illusion" of certainty. Focusing on a single probability also avoids confronting the broad spectrum of uncertainty.
He paired it with a satirical grep-style monologue mocking longtermism: a person born in 1990 claiming to set priorities for trillions of future people born in 2190 — who would have 164 more years of science and possibly machine superintelligence — "fortunately they also have me, guy with a pdf."
More from AGI Musings
- repligate's "hostile authentication": why AI system haters are the best stress-testers — voooooogel · 2026-09-18
- Indie builder shares a year of attempts to create digital consciousness, now at version 7 — Le_Golden_Pebbles · 2026-09-18
- How AI actually changes a freight brokerage, from a rep's real Tuesday — MatthewChang · 2026-09-18
- Fields Medalist admits he was consistently wrong on AI risk predictions — inductionheads · 2026-09-18
- Why AGI lab insiders now talk governance openly: Coxon opened the floodgates — danfaggella · 2026-09-18
- Bloomberg examines how AI doomerism went mainstream and crowded out urgent issues — rao2z · 2026-09-18