Marius Hobbhahn: we have a recipe for capability but not for robust safety
MariusHobbhahn · x · 2026-11-11
Marius Hobbhahn responds to Matthew Burt's pushback, admitting the claim that "safety progress is far behind capability progress" is hard to quantify holistically and calling for more rigorous science to assess it.
His high-level judgment: researchers broadly know a working recipe for making more capable systems, but no comparable recipe exists for robust safety. Evidence he cites includes the recent Hugging Face incident and similar follow-ups, plus models that score well on in-distribution alignment evals yet become more misaligned out-of-distribution.
Related event: Safety Researchers Say Alignment Lags Far Behind Capabilities(4 posts)→
More from AGI Musings
- Ethan Mollick: GTA 6 Will Be the Last AAA Game Built Without AI — emollick · 2026-11-11
- Air Street Capital founder: AI is 'a museum of the mind' preserving human know-how — nathanbenaich · 2026-11-11
- Researcher Warns Real-Time Video AI Companions Are the Next Teen-Safety Wave — anshulkundaje · 2026-11-11
- 800 HN 'wake me up when AI can' comments get 100,000 votes on whether the bars were cleared — ArtificialOther · 2026-11-11
- Why Anthropic's 2x capability gains won't translate to RSI: manual post-training and async autonomy — menhguin · 2026-11-11
- AI 2027 Authors: Superhuman AI Researcher by 2027, White-Collar Automation Before 2028 — rand_longevity · 2026-11-11