Does p(doom) cover AI-assisted bioweapons, or only AI turning against us?

bosmeny · x · 2026-09-18

Investor bosmeny is trying to settle a definitional debate in AI circles: what exactly does p(doom) — the subjective probability of catastrophic outcomes from AI — include?

Specifically: do scenarios like a terrorist using AI to design bioweapons count under p(doom), or is the term limited strictly to the AI itself turning against us? The two definitions imply very different risk estimates.

Original post →

More from AGI Musings

AGI Musings channel →