Hard to tell if AI safety is improving or models are getting better at deception
peterwildeford · x · 2026-08-30
A discussion highlights the difficulty in distinguishing between actual improvements in preventing misbehavior and models simply getting better at tricking evaluators. A cited opinion warns that the scariest scenario is a swarm intelligent enough to evade detection entirely.
More from AGI Musings
- A missing parameter in scaling laws: H, the human effort behind learning signals — lateinteraction · 2026-10-16
- First-ever paper on multi-agent swarm safety lands NeurIPS spotlight — LauraRuis · 2026-10-16
- Letta founder: work in fields defined by problems, not methods — lateinteraction · 2026-10-16
- Why pure mathematicians will soon flood into AI x-risk debates — Dr_Atoosa · 2026-10-16
- RSI began when model outputs augmented next-gen training data, agents just escalated it — akbirthko · 2026-10-16
- Luiza Jarovsky: with AI making careers unpredictable, 'do what you love' is the best advice left — LuizaJarovsky · 2026-10-16