Anthropic Reportedly Observable for Internal Thoughts
victormustar · x · 2026-07-14
This repost discusses a long article about Anthropic: the author claims that Claude's internal "thoughts" or representations can be observed via J-Space / J-lens, emphasizing that this does not prove the model has a "soul" or genuine subjective experience.
The article also notes:
- The approach is reproducible, indicating it's not pure conceptual hype
- Neuroscience researchers are cautiously optimistic about these results
- However, the relevant ablation results are insufficient to fully explain the internal mechanisms, so conclusions shouldn't be drawn based solely on them
The core message: this is a new method to observe the internal states of models, but extreme caution is still needed regarding grand conclusions like "consciousness."
Related event: Anthropic Reveals Claude's Hidden Reasoning Space: J-space(16 posts)→
More from Research
- Token Collapse: Study Finds Graph Soft Prompting Tokens in LLMs Fail to Encode Structure — Sapienza · 2026-10-27
- Terence Tao's SAIR Foundation revamps Mathathon with $1.5M compute per team — _sathvikr · 2026-10-27
- Mathathon 2.0 launches with Terence Tao's SAIR, $1.5M in compute for teams — _sathvikr · 2026-10-27
- The 'sweet lesson' is real: two kinds of ML problems need opposite strategies — mike64_t · 2026-10-27
- PixGR: Pixel-space generative decoding for fast feed-forward 3D reconstruction — kwangmoo_yi · 2026-10-27
- UniAction: unified action conditioning lets one world model handle mixed action types — ZeYanjie · 2026-10-27