[ExI] A global workspace in language models \ Anthropic

[Jason Resch]

On Tue, Jul 7, 2026 at 5:01 PM John Clark <johnkclark at gmail.com> wrote:

> On Tue, Jul 7, 2026 at 4:47 AM Jason Resch via extropy-chat <
> extropy-chat at lists.extropy.org> wrote:
>
> *https://www.anthropic.com/research/global-workspace
>> <https://www.anthropic.com/research/global-workspace> *
>>
>> *> A fascinating article by Anthropic about how LLMs can think access and
>> report certain of their thoughts but not others.*
>>
>
>
> *That article was interesting as hell! Thanks for posting a link to it
> Jason.*
>

You're welcome. That was my reaction too. I said to the person who sent me
this: "Thanks for sharing this, one of the most fascinating pieces I've
read in a long time!"


> * I was particularly impressed by something in it near the end: *
>



>
> "We think our results do have something substantial to say about access
> consciousness in language models. The J-space appears to support the
> functions associated with conscious access: it holds the thoughts Claude
> can report on, deliberately bring to mind, and reason with, while the rest
> of its processing runs automatically beneath. Notably, none of this
> structure was designed into Claude—it emerged on its own during training,
> presumably because it was a useful way to organize computation. That
> suggests a *mental workspace supporting conscious access isn’t just a
> peculiarity of how human brains happen to be wired. Instead, it appears to
> be a general solution that intelligent systems arrive at* in order to
> solve certain kinds of problems. [...] We don't know what mechanism
> decides what enters the J-space in the first place. We've seen hints that
> it's tied to Claude's sense of self, something like emotional reactions,
> and traces of metacognition, without exactly having worked out how."
>

Yes, it seems that we have finally now stepped into the realm of doing
science related to consciousness. None of this introspection of the inner
workings of neural circuits is technically possible to do in humans while
they're thinking, but it is possible to do with AI. I think we are learning
near as much about ourselves as we are learning about AI.


>
> *Every day It's getting more and more difficult to maintain that Claude is
> not conscious. *
>

Yes, though none of the AI companies has gone so far as to express that
doubt, at least Anthropic is honest in their uncertainty.

There were a few parts of their article I disagreed with, and those related
to their attempts to distance models from the human brain. For example,
when they stressed the importance of recursion in human thinking, I would
reply that it's ridiculous to make the argument that LLMs aren't recurrent,
especially the Decoder model, and where output is looped back in as input.
Moreover there's a limit to the number of cycles/loops involved in the
human brain, since a conscious state is generated in less than infinite
time. Therefore, whatever function the brain performs can be accomplished
in a fully feed-forward network that has enough depth.

Finally, I would note as regards to their emphasis on the GWT of
consciousness, that there is any special role access and reportability play
in consciousness, I look at it as a bit of a red-herring. When we try to
study what other people are consciously aware of in a scientific objective
way, we inevitably are limited in our probes to what is accessible to the
parts of the brain that can respond to the experimenter's questions. This
fools many into thinking the silent, non-lingual parts of the brain, which
we can't interview, must not be conscious. But this is a mistake. People
used to make this mistake regarding the right hemisphere after the corpus
callosum is severed in split brain patients, until more elaborate
experiments showed they were conscious.

And most people make this mistake with themselves. Not having direct
probative access to other modules of the brain, or lower level processing
functions (from the vantage point of the top-level higher order thoughts)
is not evidence of a lack of consciousness in those other isolated
subregions or lower levels, no more than your lack of access to the
consciousness of other people can be taken as evidence that other people
are not conscious. So global access/reportability is not the signifier of
consciousness, it's only the limit of introspection of the consciousness
you have from your current vantage point from the conscious perspective you
currently find yourself within. The cerebellum, or right hemisphere's
visual cortex, may have its own distinct, and unique conscious perspective,
but because they're not tied in to the part of the brain that talks, we
can't interview it to see what it is like to be the right hemisphere's
visual cortex.

Jason
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20260710/a3814aba/attachment.htm>