ChinaTalk ModelTalk: Why the Hugging Face hack broke through and AI safety's new coalition
ChinaTalk · rss · 2026-09-15
ChinaTalk's ModelTalk with Nathan Lambert and Jasmine Sun covers a pivotal week: Dario asked the industry to slow down while Trump called it a conspiracy. Key threads: the Hugging Face hack broke containment because it pattern-matches forty years of rogue-AI sci-fi (Terminator on the NY Post cover), colliding populist data-center anger with frontier safety concerns into a mega regulatory moment; Bernie Sanders did more than anyone to convince the left AI risk is real, enabling an unlikely coalition with Tucker Carlson and Rogan; Nathan Lambert argues "nobody can measure the frontier" makes pacing beat pause; plus METR's trust monopoly, CAISI's $10M, whether China (Zhipu, Moonshot) will slow down too, and military secrets leaking into Claude via proxy routers.
More from AGI Musings
- Critic warns against conflating vibes-based AI risk estimates with empirical likelihoods — merrierm · 2026-09-15
- whurley mocks AI doomerism: 4% unemployment, no evidence for the job-loss thesis — whurley · 2026-09-15
- Anthropic's worst-case AI scenario: US economy grows 32% by 2030, but almost none of it goes to you — ahuja_priyank · 2026-09-15
- Alsop: fear-driven AI regulation advocates all have a stake in regulatory capture — StewartalsopIII · 2026-09-15
- p(doom) is vibes, not statistics: researcher offers a VET framework for AI discourse — merrierm · 2026-09-15
- DeepSeek's Liu Shengyu clarifies viral post: a farewell to hand-written kernels, not job anxiety — teortaxesTex · 2026-09-15