Venkatesh Rao essay: EAs are great at numerical uncertainty, bad at deeper doubt

binarybits · x · 2026-09-22

In the essay 'EA Safety' on Contraptions, Venkatesh Rao argues that Effective Altruists are 'good at expressing uncertainty numerically while being remarkably bad at sustaining deeper forms of doubt about the conceptual machinery within which those numbers are generated.' Drawing on his 20-year relationship with the rationalist community — which grew from the Overcoming Bias blog into an institutional force shaping AI's future — Rao, labeled 'postrationalist' by Scott Alexander, suggests EA's claim to solve AI Safety creates two problems, not one.

Original post →

More from AGI Musings

AGI Musings channel →