A Drake Equation for AI extinction risk decomposes the debate into six conditional probabilities

Mazzaroth · reddit · 2026-09-14

A Reddit user proposes decomposing AI extinction risk using a Drake-equation-style chain of conditional probabilities: superhuman AI (0.8) × dangerous deployment (0.7) × serious misalignment (0.3) × escape from control (0.5) × durable strategic control (0.5) × extinction given control (0.4), yielding a rough P(X) ≈ 1.7%.

The author stresses the numbers are deliberately debatable—that's the point. Decomposition turns "Will AI kill us all?" into a series of more workable questions, and reveals that people quoting 0.01% vs 20% extinction probabilities may only disagree on one or two links in the causal chain.

Original post →

More from AGI Musings

AGI Musings channel →