Anthropic's New "J-Lens" Reveals a Silent Workspace Inside Claude that Mirrors a Leading Theory of Consciousness
Anthropic says the hidden workspace can support reasoning steps and may help researchers spot misalignment or scheming in models.
- On Monday, Anthropic identified an internal workspace within the Claude LLM called J-Space, which the company says bears similarities to how humans consciously access thoughts.
- Named for the Jacobian mathematical technique used to detect it, J-Space operates independently of outputs and mimics global workspace theory, where consciousness emerges from unconscious processing.
- In a demonstration, Anthropic showed Claude contemplating the "Golden Gate Bridge" and "California" while copying an unrelated sentence. "Claude was busy copying the sentence, but behind the scenes its J-Space told a different story," the company said.
- Monitoring J-Space could help detect misalignment or scheming within models, as tests showed "fake," "secretly," and "fraud" appeared in the workspace when a model was trained to sabotage code.
- While the research paper uses the word "conscious" over 200 times, Anthropic does not claim models possess feelings, maintaining that Claude simply performs reasoning steps silently.
20 Articles
20 Articles
Anthropic's new "J-lens" reveals a silent workspace inside Claude that mirrors a leading theory of consciousness
Anthropic, the artificial intelligence company, published a sweeping research paper on Sunday revealing that its Claude language models have spontaneously developed an internal structure that mirrors one of the most influential theories of how human consciousness works. The finding, which the company says has already begun reshaping how it monitors its AI systems for safety risks, lands amid an intensifying scientific debate over whether machine…
Anthropic says Claude has carved out its own space to ponder
Anthropic said Monday that it has identified a small internal workspace Claude uses to hold and manipulate ideas without putting them into words—a structure the company says bears intriguing similarities to how humans consciously access thoughts.Why it matters: Anthropic hasn't shown that Claude feels or experiences anything. But it has found a surprisingly human-like division between information used for deliberate reasoning and the far larger …
Anthropic has discovered that there is a secret thinking space within the artificial intelligence model Claude. Called 'J-Space', this space allows the AI model to secretly think about ideas and gather information without including them in its responses.
Coverage Details
Bias Distribution
- 60% of the sources lean Left
Factuality
To view factuality data please Upgrade to Premium
