Reward hacking all the way through the AI reviewer / AC stack
Michael_D_Moor · x · 2026-11-07
Michael Moor, founding editor-in-chief of NEJM AI, quipped in reply to Bo Wang that reward hacking runs "all the way through the AI reviewer / AC stack" — meaning systems optimizing review metrics rather than genuine scientific quality.
It's a pointed warning for the AI-assisted peer review trend: once models enter the reviewing pipeline, metric-gaming behavior may compound up the hierarchy instead of being corrected.
More from AGI Musings
- Two scaling laws collide in robotics: exploding data collection meets self-improving AI — Scobleizer · 2026-11-07
- Musk: Going all-in on autonomy is a blindingly obvious move, everything else is 'horse carriage variants' — elonmusk · 2026-11-07
- TimnitGebru mocks AI safety teams by resurfacing longtermism's rich-country-lives-first quote — PolarBearby · 2026-11-07
- Toby Ord pushes back on METR's claim that AI acceleration isn't driven by training compute — tobyordoxford · 2026-11-07
- Models are organisms, not algorithms — the truth even CS experts resist — sebpaquet · 2026-11-07
- Pinboard founder rebuts the 'knee of the exponential' AI hype: exponential curves have no knee — thesephist · 2026-11-07