Berkeley-led report argues a hardwired pause of frontier AI training is feasible, challenging Hinton
wfithian · x · 2026-10-24
A new working-group report, "The Feasibility of a Hardwired Pause of Frontier AI Training," organized by UC Berkeley statisticians William Fithian and Wesley Holliday with contributors from Oxford, Princeton, Stanford and Mila (including Barry Eichengreen and David Krueger), argues frontier AI training could be halted by controlling chip supply. Fithian publicly challenged Geoffrey Hinton, citing earlier conversations where Hinton seemed to agree chips could be shut off — prompted by Hinton's latest pessimism in his Gödel Lecture at TU Wien.
Related event: Scholars Argue a Hard Pause of Frontier AI Training Is Technically Feasible(4 posts)→
More from AGI Musings
- Cheap code isn't enough — institutions won't rewrite their stacks until forced — jd_pressman · 2026-10-24
- AI safety debate: those right on AI cyberattacks may be wrong on bioweapons — basedjensen · 2026-10-24
- Parent uses 214-revision document history to overturn daughter's false AI-cheating accusation — AlexTensor · 2026-10-24
- Flash-to-Silverlight playbook: AI may force a rewrite of software defaults — jd_pressman · 2026-10-24
- AI in bio researchers warn: too many tool builders and benchmarks, too little real science — anshulkundaje · 2026-10-24
- Anna Salamon floats taking down the Internet if AI hardware bugs prove severe — jd_pressman · 2026-10-24