AI safety researcher slams community for ignoring where CoT monitoring trajectory leads
JacquesThibs · x · 2026-11-05
Jacques Thibs cites a finding that models' decisions to underperform often weren't verbalized in chain-of-thought, meaning CoT-based misalignment monitoring failed to detect them. He argues the AI safety community keeps funding "the next step on the trajectory" instead of looking at where this all leads — a structural blind spot in how safety research resources are allocated.
More from AGI Musings
- Many grant solicitations now flagged as 100% AI-generated by detector Pangram — analisereal · 2026-11-05
- Why AI research keeps looking under the lamppost: stale problem frames inherited from Kant — fkasummer · 2026-11-05
- Michael Nielsen on why scientists can't tell an anomaly from a revolution — michael_nielsen · 2026-11-05
- Anthropic's S-1 Paints an Autonomous Labor Future—and an Oversight Problem — asusarla · 2026-11-05
- New paper uses 288M job postings to measure where AI raises the value of work it can't do — robseamans · 2026-11-05
- Jeff Ladish: Claude Is a Category of Models, Not a Single AI — JeffLadish · 2026-11-05