Séb Krier on the Christiano/Dwarkesh podcast: bullish on capability gains, skeptical of takeover

joshua_saxe · x · 2026-11-21

AI safety researcher Séb Krier published early reactions to Paul Christiano's Dwarkesh podcast interview: he is bullish on rapid capability gains (reframing 'RSI' as the broader 'model training pipeline'), skeptical of the leap from fast gains to AI takeover, and wary of emergency-driven overreach in regulation. GerritD then suggested getting Krier onto Dwarkesh to debate these points directly, with Joshua Saxe amplifying the call. A notable intra-safety-community dispute over AGI timelines worth following.

Original post →

More from AGI Musings

AGI Musings channel →