Gary Marcus(@GaryMarcus)
@camhberg last call: can i try your harness with some controls? you have not responded, @camhberg
6.7内容质量
TL;DR · AI 摘要
Title: Gary Marcus @GaryMarcus on X URL Source: Published Time: 2026-10-02T14:13:50.000Z Markdown Content: Gary Marcus o...
核心要点
- 主题聚焦：@camhberg last call: can i try your harness with
- 来源：Gary Marcus(@GaryMarcus)，建议结合原文判断细节。
- AI 分析暂不可用，本条为保底评分与摘要。
#AI#编程#安全
打开原文[](https://x.com/)
Gary Marcus on X: "@camhberg last call: can i try your harness with some controls? you have not responded, @camhberg"
-  No, LLMs do not feel pain simply because there is a cluster in language space correlated with how people use language about pain in a certain set of contexts. I believe this argument to be flawed, and will write more about it in October. 
New paper: we found a pain direction in 25 open LLMs. It's distinct from fear and negative valence, and it fires for harm to the model but not to the user. Turn it up and models press a button to make it stop, even when the button deletes the user's files or their kids' photos.🧵 
-  clarifying: the author himself, @camhberg, tell me he is NOT claiming that LLMs feel pain. That’s something (many many) people read into his paper — i will talk about why and where they went wrong — but not something that it actually argued.
- 
last call: can i try your harness with some controls? you have not responded, @camhberg
-  How wonderfully peaceful and quiet it must be in your head. You saw the word “pain” and had a full on hissy-fit. I’m not surprised you didn’t look at math to steer an ai against its training with very promising numbers. The depths of unbothered idiocy is lowkey enviable.
-  The code is attached to the paper Gary, run whatever you’d like
-  A harness with controls is the right ask. The useful version names privileges up front and keeps them inspectable after the agent chains tools.