Casper: 'more evals' is a weak pacing lever, implementations range from safety-washing to real oversight
StephenLCasper · x · 2026-09-23
In his thread on embedded evaluations, Stephen Casper argues: (1) calls by Dario and others for embedded evaluation have not been paired with stricter pacing measures like hardware controls, making 'more evals' a remarkably weak mechanism for slowing AI progress; (2) embedded evaluation spans a huge range — from net-harmful safety washing to rigorous oversight where auditors can access records, data, models and meetings, and whistleblow.
Related event: Researcher criticizes embedded evaluations as weak pacing substitute(2 posts)→
More from AGI Musings
- Juan Benet podcast reading list: The Beginning of Infinity, Permutation City, Nexus and more — juanbenet · 2026-09-23
- "Slowdown talk is aging terribly" — new models keep dropping, curve bends harder — Dr_Singularity · 2026-09-23
- Palantir co-founder Joe Lonsdale says everyone will be 'really wealthy' in the 2030s — Polymarket · 2026-09-23
- Goodside reconsiders anti-pause stance after labs call to slow the frontier — goodside · 2026-09-23
- Robinhood's Vlad Tenev: AI Will Create More Lawyers and Engineers, Not Fewer — PeterDiamandis · 2026-09-23
- NumPy creator Travis Oliphant: open abstractions beat AI vendor lock-in — teoliphant · 2026-09-23