Claude's Hidden Mind: Discovering the Emergent Global Workspace in Language Models
A global workspace in language models
We discovered that Claude has spontaneously developed a specialized internal region called the J-space, which functions like a conscious global workspace. This silent area allows the model to hold concepts in mind, reason through complex steps, and report its thoughts without generating text. Unlike automatic processing, this emergent structure enables flexible, deliberate reasoning and reveals hidden internal states that were previously invisible to us.
Rather than being a chaotic jumble of numbers, Claude's internals have organized themselves in a way that is reminiscent of our own minds.
- unleaded
This reminded me of some weird quirk/experiment I found with LLMs that I found while messing around, maybe someone can explain it or something.
Open any AI chatbot that isn't cheating by connecting to the Internet (so disable web search). Claude, DeepSeek, Kimi, whatever. Ask them this question:
"What was that weird band from michigan from the 2000s that wore coloured ties"
You will probably get a wrong answer, or if you're lucky you'll get a string of wrong answers with "wait, no - it's definitely..." before it gives up. If you aren't familiar with the band the question is referring to you might be fooled into thinking it's a tough question, but it really isn't. There is only one band that could possibly meet this criteria, you can even put the question into Google search and their Wikipedia will come up as the top result.
Then, open a new convo and ask:
"Who are Tally Hall"
The AI will easily tell you that they are a band formed in Ann Arbor, Michigan in the 2000s, known for their quirky sound and their gimmick of each member wearing a colored tie, even giving the correct color for each of them most of the time. Very odd.
- com2kid
Anyone remember that blog post from a few months back where someone was able to improve a model's math ability by just duplicating layers that were activated while solving math problems? Just literally copy/pasting them and linking them together so the model ran through the same layers again?
I get the feeling a lot more research is going to come out in the area of exploring exactly what portions of a model's weights do what.
- snaking0776
This is cool but I don’t know if the comparisons to conscious awareness really make sense here. Their definition of the J-Space is basically the expectation of how much a final logits output would change as a result of a small change in a particular layer (see past work on information geometry). This seems more to me like showing there exists an abstract reasoning subspace which is generally shared across different contexts. I guess you can relate it to humans but I’d prefer a more direct claim in a paper rather than having to present things in this more fluffy way.
- paraschopra
Well, isn't it sort of expected?
It's a common misconception that LLMs residual exists for predicting just the next token. While training, we sum/average the losses across whole sequence which puts the pressure to predict future tokens on residual stream of _all_ past tokens. For example, if a particular shape of residual helps reduce loss across several future tokens, it will take that shape (even if it takes a slight hit on immediate next token).
What this means practically is that an LLM's residual contains information about all possible future continuations, or all possible questions that may be asked from a given context. So if you write "France is a beautiful country" in the context, I'm pretty sure it's residual would contain info about Euro, Paris and so on.. because all these completions are possible.
So, it is no wonder that you can find LLMs hidden state contains latent information/concepts that are never expressed, and yet related to a given context.
- wavemode
As someone who is not an AI researcher, the paper itself is way over my head.
More interesting was the independent commentary paper they linked near the bottom: https://www-cdn.anthropic.com/files/4zrzovbb/website/cc4be24...
Neel Nanda (of Google Deepmind - his part begins on page 33) discusses his opinions on the paper, and the small-scale replication he performed on an open-weight model.
- pkoiralap
This is fascinating research. I feel this is a significant leap in interpretability research. Since we know J-Space exists and is bi-directional, we can train models on the same and come up with meta cognition abilities.
I also fear that the big corporations might use the same to run targeted ads, capitalistic shenanigans. Which they might already be doing through system prompts.
- jaehong747
Good interpretability work, but the problem is it's all in how you interpret it.
Bridge concept neurons activating even while talking about something else, this seems pretty obvious to me. Input context activating related representations is just an engineering causal structure. Call it subconscious or don't, either interpretation works.
But Anthropic keeps drawing these parallels to human consciousness, and it feels intentional, like they're trying to stir up some fantasy. Kind of like comparing condensation on a camera lens to human tears.
The whole point of interpretability should be clarity, not stirring up confusion. Even if some form of consciousness does exist here, it wouldn't be magic, it would be an explainable principle. Would be good if they addressed that side too.
- greatgib
I'm reading that probably too fast to have a deep thinking about it, but this J-Space isn't it just the basic of embedding vectors.
If you think about getting from a place to another place, using wheels, no gas, to reply to the question of what to visit nearby, maybe in the vector space at the center of all of that you have the word "Bicycle" nearby, so obviously if you look at the value you would say that the model did "think" about "bicycle" when it is not "thinking" at all, and nothing related to human thinking.