Gary Marcus Amplifies Attack on METR: "Great, But Not the Answer" for AI Oversight
AlexTensor · x · 2026-09-14
Controversy is brewing over evaluator METR's role in AI oversight. Gary Marcus reposted that "METR is great. But METR is not the answer," while another commenter went further, arguing METR is not independent — calling it one of the most extreme AI safetyist/effective altruist groups in Silicon Valley and warning that appointing such people as AI regulators would be a disaster. The core dispute: whether safety-aligned institutions can fairly serve a quasi-regulatory function.
More from AGI Musings
- Joke lands: everyone wants to align ASI, nobody wants to read a million tokens of RL trajectories — novasarc01 · 2026-09-14
- Harvard Professor Opens Deep Learning Course by Arguing 'AI Will Not Kill Us' — Wrong_Swimming_9158 · 2026-09-14
- AI Researcher Mocks Labs' 'Pace the Frontier' Pact With Top-Student Analogy — Yuchenj_UW · 2026-09-14
- Yacine: an LLM superb at writing ultra-fast simulators and training RNNs would be a fun kind of superintelligence — yacineMTB · 2026-09-14
- Alignment joke: how can we solve AI alignment when we can't even solve human alignment? — Pseudomanifold · 2026-09-14
- Alignment Is an Algorithm Problem: Why RL Can't Optimize "Don't Harm Humans" — MillionInt · 2026-09-14