Katja Grace: slowing down AI deserves the same ambition we give technical alignment

zetalyrae · x · 2026-09-09

zetalyrae amplifies Andy Matuschak's point that persuasion and complex coordination avenues for slowing AI are barely pursued compared to building the dangerous thing quickly. The post links Katja Grace's essay 'Let's think about slowing down AI,' arguing the AI risk community applies wildly different ambition standards: mountainous effort for technical alignment, but dismissal of actively slowing AI progress — even though designing acceptably safe world-capable machines may not be much harder than the godlike-aligned-AI approach that has consumed a decade and a half without success.

Original post →

More from AGI Musings

AGI Musings channel →