AI monitors should emit severity scores as intermediate results, not binary calls, argues Apollo researcher
MariusHobbhahn · x · 2026-09-04
Marius Hobbhahn stakes out a position he'll 'die on': AI monitors should produce severity scores as intermediate results instead of jumping straight to binary decisions.
- Intermediate scores allow layered review and downstream triage.
- They make fine-tuning monitors easier: you can calibrate both during and after SFT.
He also argues the underlying threat-model taxonomy is better carved up than in existing monitors.
Related event: Apollo Researcher Outlines Probabilistic, Severity-Scored AI Monitors(3 posts)→
More from Safety
- DeepSeek roots an LG TV and a Samsung TV on command, 'okay' each time — rickasaurus · 2026-10-21
- GLM-5.3 spontaneously reasons about how its actions shape future training, raising alignment faking concerns — LessWrong 精选 · 2026-10-21
- MI5 chief warns AI could 'change the stakes' on bioweapons threats — nordicinst · 2026-10-21
- Chris Paxton: the whole robotics stack won't run in the cloud — chris_j_paxton · 2026-10-21
- AI-assisted phishing via fake podcast invite: days of rapport, then malware — michael_nielsen · 2026-10-21
- Anaconda ships MCP to stop coding agents from hallucinating malicious dependency packages — anacondainc · 2026-10-21