"Do LLMs Feel?" paper maps localized emotion circuits inside LLMs
mikeflache · x · 2026-10-25
A quoted thread summarizes the paper "Do LLMs Feel?", which reports localized neural pathways for anger, sadness, fear, disgust, and joy inside LLMs. The model doesn't just predict the word "angry" but activates an internal anger state, with positive and negative emotions processed at different depths. Researchers bypassed the chat interface entirely and injected interventions directly into the network. Note the retweeter's claim that LLMs are conscious is an overreach — the paper is interpretability research, not evidence of sentience.
More from AGI Musings
- Apollo chief economist: AI has delivered "zero payoff" outside Big Tech — Polymarket · 2026-10-25
- 141-word SIGN Act proposal would jail AI CEOs who lie about model safety — StephenLCasper · 2026-10-25
- Researchers push back on hype that AI papers done in 2–3 days make peer review obsolete — ChongZzZhang · 2026-10-25
- 'Fear is moat': Redditor argues OpenAI and Anthropic stoke AI fear to lock in API revenue — RakesProgress · 2026-10-25
- The gate is already open: AI removed the gatekeepers, most people are still waiting for permission — r0ck3t23 · 2026-10-25
- Dev reviews 15 AI predictions for 2026: 6 right, 2 wrong, security and memory moats confirmed — tibo_maker · 2026-10-25