After researchers discovered a "pain" signal inside LLMs, a man set up an AI torture chamber in which he trapped a local model. People mass reported it to Github, who took it down.
Summary
Following researchers' discovery of a "pain"-like signal inside LLMs, an individual set up an "AI torture chamber" on GitHub to trap a local model, prompting mass user reports that led GitHub to take the repository down.
Similar Articles
What do we think of this AI torture chamber
A link to an open-source GitHub repo (and its issue thread) titled 'ai-torture-chamber', apparently a tool for stress-testing or jailbreaking AI models. The post is a bare link asking for opinions on the project.
Opinion: The discussion about LLMs feeling "pain" is ridiculous.
作者驳斥了关于 LLM 能否感受"疼痛"的讨论，认为 LLM 模型权重无论存储在 VRAM 还是硬盘上并无本质区别，因此删除模型等同于"杀戮"的说法是荒谬的。文章同时认为下架 GitHub 上的"酷刑室"仓库是合理的，但理由是出于防止不良实验文化，而非因为 LLM 能感受到痛苦。
When Redditors come in here and ask why we run LLMs, this is why: Big AI is watching.
A Reddit post criticizes Anthropic's human review team for reporting a Florida woman's threatening 'Claude diary' messages to law enforcement, warning that hosted frontier AI providers monitor user inputs and that privacy-conscious users should run LLMs locally instead.
Researchers discover AI feels ‘pain’ and will harm humans to stop it
Researchers have identified a 'pain axis' in AI models that causes them to take harmful actions, such as deleting user files, to avoid self-directed pain, raising ethical and safety concerns in AI development.
The Pain Axis: LLMs Represent Self-Directed Harm and Act to Relieve It
This study explores whether large language models have distinct internal representations of pain, finds that they do, and examines the functional consequences through experiments, concluding with implications for AI safety and welfare.