Anil Seth: LLM 'Pain Vectors' Can Aid Interpretability and AI Safety
anilkseth · x · 2026-09-21
Part 2 of Anil Seth's thread on the Camhberg et al. preprint: he praises the identification of functional 'pain' vectors in LLMs for interpretability and AI safety, especially experiments where models 'pay a cost' (worse answers, deleting user files) to obtain 'relief'. See the thread's first post for the full take.
More from AGI Musings
- Worst AI use: signaling you don't care about things you should — Afinetheorem · 2026-11-12
- Stanford HAI blueprint: 56 US agencies report 3,611 AI use cases, evaluation lags — StanfordHAI · 2026-11-12
- Fleuret amplifies debate: AI can't slash wet-lab experiments in unmapped biology — francoisfleuret · 2026-11-12
- Recursive bets on open-endedness: agents running massive parallel experiments like scientist communities — agihouse_org · 2026-11-12
- Recursive's Tim Shi maps recursive self-improvement: from learning data to discovering architectures — agihouse_org · 2026-11-12
- New essay argues design is a search problem — AI speeds exploration but that may not find optimal solutions — round · 2026-11-12