26 researchers propose a verifiable global pause on frontier training via chip supply cutoff
wfithian · x · 2026-10-23
In a debate over who bears moral blame for AI misuse, statistician Will Fithian pushed back on the claim that hackers mostly use Chinese open-weight models so US labs shouldn't be blamed. He argued those models are distillations of OpenAI/Anthropic/GDM capabilities, built on methods—chain-of-thought, RLHF/RLVR, Chinchilla scaling, sparsely gated MoE, and the Transformer itself—developed largely by US frontier labs and researchers.
He further argued all such models depend on AI accelerator chips, whose exponential proliferation governments could cut off globally once they decide to. He and 26 co-authors wrote up how governments can implement a lasting, verifiable, international pause on frontier training.
Related event: Scholars Argue a Hard Pause of Frontier AI Training Is Technically Feasible(4 posts)→
More from AGI Musings
- Dario's predicted timeline distilled: $1B one-person company in 2026, AI beating Nobel winners by 2027 — AccBalanced · 2026-10-24
- VC: X feels saturated but enterprise CEOs still aren't using AI — it's so early — karinanguyen · 2026-10-24
- AI founders argue the digital ecosystem needs rebuilding from scratch — creating jobs for now — max_paperclips · 2026-10-24
- New paper introduces "bottleneck pressure" to find where AI raises the value of work it can't do — soumitrashukla9 · 2026-10-24
- Meme/persuasion risk is more compelling than bio risk, argues teortaxesTex — teortaxesTex · 2026-10-24
- Why the 'AI creates 8x more jobs' claim doesn't hold: unpacking the U.S. Chamber survey — BuddyWeary653 · 2026-10-24