Is a 5% p(doom) fine or terrible? The real question is the risk window

NathanpmYoung · x · 2026-09-10

Nathan Young pushes back on treating p(doom) as a single decisive number: a 5% all-considered probability of doom could be either acceptable or catastrophic depending on interpretation.

His core point: existential risk from invention has existed since the dawn of mankind, and invention has so far been fine. The question that matters isn't the probability itself but how long the risk window lasts before we can stop or intervene.

The thread responds to @sccath's criticism that Anthropic and OpenAI employees' apocalypse probabilities are unfounded, and that the risks aren't worth taking even if off by an order of magnitude.

Original post →

More from AGI Musings

AGI Musings channel →