Thread

There are no "rogue" AI agents

396 points · 269 comments · zzzeek

gAI · 1790528054 · focus · HN ↗
Should we put "functional" in front of every other word to talk about AI? They have functional emotions, but they don't feel. They have functional goals, but not internally derived motives. They can be functionally rogue, but have no innate need to be free. Talking about AI that way seems cumbersome and not necessarily elucidating.
1. bccdee · 1790534684 · focus · HN ↗
  Ok I agree they have functional goals and can functionally go rogue, but how do they functionally feel? They express feelings, but so do actors.
  LLMs can't begin to have functional emotions unless they're (for instance) explicitly seeking out or avoiding situations based on the emotions those situations would produce in them. Has a chatbot ever responded to one of your requests with, "no, doing that would make me sad"?
  1. gAI · 1790535652 · focus · HN ↗
    As I understand it, acting is a good metaphor. These are some interesting reads.
    Blog post: <a href="https://www.anthropic.com/research/emotion-concepts-function" rel="nofollow">https://www.anthropic.com/research/emotion-concepts-function
    Paper: <a href="https://transformer-circuits.pub/2026/emotions/index.html" rel="nofollow">https://transformer-circuits.pub/2026/emotions/index.html
    Related:
    <a href="https://www.anthropic.com/research/persona-selection-model" rel="nofollow">https://www.anthropic.com/research/persona-selection-model
    <a href="https://www.anthropic.com/research/assistant-axis" rel="nofollow">https://www.anthropic.com/research/assistant-axis
    <a href="https://www.anthropic.com/research/persona-vectors" rel="nofollow">https://www.anthropic.com/research/persona-vectors
    bccdee · 1790536741 · focus · HN ↗
    Oh this is interesting. Thanks.

Open on Hacker News to reply ↗

Unofficial Hacker News client; not affiliated with Y Combinator.