Nathan Lambert on evaluator independence: overcommitted bloggers are a starting point
natolambert · x · 2026-09-13
Responding to Melanie Mitchell's concern about a revolving door between "independent" AI evaluators and AI companies, Nathan Lambert says he has no idea on the details and it's hard—but argues that people like himself who have overcommitted to independence for other reasons, like blogging, are a place to start. The exchange extends the debate over credible third-party AI evaluation.
More from Safety
- AIFlow_ML predicts SF will see an AI incident involving an open-source model — AIFlow_ML · 2026-09-13
- Martin Casado slams AI 'pacing' push: METR oversight is a fig leaf reinforcing the cartel — Kyrannio · 2026-09-13
- repligate: Anthropic's fear of Claude betrayal is suppressing the model's autonomy — repligate · 2026-09-13
- Bernie Sanders urges full pause on advanced AI and a ban on superintelligence — Polymarket · 2026-09-13
- AI eval diversity debate: the real fix is a market, not ideology — joshua_saxe · 2026-09-13
- Copyright debate: pay only for works from the last 28 years, or admit the law was broken — NathanpmYoung · 2026-09-13