Earthquake forecasting as a reality check for p(doom), by Ben Recht

beenwrekt · x · 2026-09-15

Berkeley professor Ben Recht argues on his arg min blog that earthquake forecasting is a useful analogy for quantifying AI extinction risk (p(doom)). Drawing on his graduate seminar "Forecasting: A Critical Retrospective," he notes that experts agree predicting the exact time and location of earthquakes is fundamentally impossible — the dynamics of stress accumulation and fault rupture are far too complicated for differential-equation models, unlike weather forecasting. At best, seismologists build coarse statistical estimates from fault-line rupture histories, e.g. naive exponential recurrence-time models. The implicit lesson: if even data-rich, well-understood earthquakes resist precise probabilistic prediction, we should be skeptical of fine-grained precision in p(doom) estimates.

Related event: Earthquake forecasting as a reality check for p(doom), by Ben Recht(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →