Others are about what LLM minds might be like (by @winniestreet and Geoff Keeling, Jon Simon, and Helen Yetter-Chappell). And the last two are about what it even means to ask whether AI systems are conscious (by Tim Bayne and Geoff Lee).
A whole issue of new papers on AI minds and consciousness! Jonathan Simon (@futureofcitizenship), @dcshiller and I have edited a special issue of the Journal of Consciousness Studies on 'Consciousness in Current AI', and we're delighted about how it's turned out. Link below.
So this new research might be evidence that LLMs have welfare-relevant properties other than p-consciousness. We've got a lot of scientific and philosophical work to do to figure this out, but I'm excited about this step forward.
Second, I don't think it's clear that phenomenal consciousness is necessary for being a welfare subject. Other minds might have consciousness-like properties that are sufficient. For this kind of view, see e.g. philarchive.org/rec/LEECPA-5.
I noticed some people reacted to this work saying "it's not consciousness." Of course, there is always difficulty in claiming consciousness in AI, and Anthropic is very careful about discussing it.
Actually, I think the future of consciousness science should look more like this. Less mystery, more mechanisms. The approaches like J-Lens and J-Space are very useful when we think about access, reportability, control, and reasoning in human neuroscience.
If consciousness is ever demystified, it will be through work like this.
We should be careful not to put too much weight on the phenomenal/access distinction; we don't know how p-consciousness is related to access and other functional phenomena.
First, even assuming that phenomenal consciousness is distinct from access, evidence of a workspace is evidence of phenomenal consciousness. Maybe what p-consciousness requires is a workspace plus some enabling conditions.
This is why thinking about the enabling conditions of consciousness matters so much (as opposed to just focusing on what theories of consciousness say). See my new preprint on the topic: philarchive.org/rec/MICOCA-3.
Can you expand on why it is argued that this research is welfare-relevant? I don’t see what supports the ideas in the fifth bullet point, given that this research doesn’t say much about phenomenal consciousness or sentience…
As @rgblong has said, we wrote a commentary on this research (with @dcshiller and @dillonplunkett). Part of our intention was to put an accessible overview and discussion in one place for those whose interest is mostly in consciousness and welfare. Link and comments below:
New Anthropic research: A global workspace in language models.
Of everything happening in your brain right now, only a tiny fraction is consciously accessible—thoughts you can describe, hold in mind, and reason with.
We found a strikingly similar divide inside Claude.
I’m mentoring Autumn 2026 @MATSprogram Fellows interested in doing AI welfare research.
The application deadline is this Sunday (6/7). More info in this thread:
We're hiring Research Scientists to join my team at @eleosai!
We do foundational and applied ML research on the moral status and potential well-being of AI systems.
This is urgent, important work, and Eleos is an extraordinarily fun and exciting place to do it.
Details below.