Timothy B. Lee: mid-term AI forecasting is overrated for policymaking — wait and see works

binarybits · x · 2026-09-19

Timothy B. Lee argues people overestimate the value of medium-term forecasting for AI policymaking: if something weird is 5 years out, doing nothing makes sense since you'll have more information next year. Even if it starts next year, gradual development means little downside to making policy after it begins. He adds this doesn't hold under a "fast takeoff" — which he doesn't believe, but many AI safety people do.

Related event: Dean Ball warns of userless agents humans can't switch off(4 posts)→

Original post →

More from AGI Musings

AGI Musings channel →