On lab pacing agreements: taming superintelligence is natural science, not pacing deals

tszzl · x · 2026-09-13

@jachiam0 argues lab cooperation on pacing frontier automated AI research is net good but won't much reduce risk — estimates of p(good future) or p(doom) shouldn't swing much. tszzl replies that whether labs can tame superintelligent models will mostly be a matter of natural science rather than pacing agreements.

Related event: OpenAI researcher: taming superintelligence is science, not pacing deals(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →