Pedro Domingos: We need Goodhart-proof measures for AI evaluation
pmddomingos · x · 2026-09-03
Renowned computer scientist and author Pedro Domingos posted a terse take: "We need Goodhart-proof measures."
The one-liner targets a core pain point in AI evaluation: once a metric becomes an optimization target — like benchmark leaderboards — over-optimization destroys its meaning, per Goodhart's law. It's a direct jab at the credibility of eval-centric model launches.
More from AGI Musings
- Cambridge professor David Krueger slams Dean's apology: 'You don't get to just say my bad for deceiving you' — DavidSKrueger · 2026-09-03
- Anders Sandberg to speak on human autonomy in the AI age at EAGx Oxford, Sept 25-27 — anderssandberg · 2026-09-03
- Using OpenEvidence, a user found a cancer clinical trial that saved his father — saranormous · 2026-09-03
- Gary Marcus mocks Sam Altman's AI bubble warning: bubble architect calls it a bubble — GaryMarcus · 2026-09-03
- CoT monitorability not abandoned yet, but new techniques risk a race to the bottom — DavidSKrueger · 2026-09-03
- Sam Altman warns at G20: cybersecurity things 'will go very wrong' without urgent action — RebeccaBellan · 2026-09-03