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:

Bradford Saad, co-author of the new CMEP-Eleos report "Studying AI Welfare Empirically," wrote an essay summarizing the report and offering key takeaways from his perspective. Highly recommended! meditationsondigitalminds.su…

Eleos wrote a commentary Tl;dr -important, excellent work -we’re more cautious than the authors about the stronger claims of 'global workspace' -still, it's evidence in the direction of access consciousness -investigating AI consciousness is tractable and urgent More:

Is AI close to being conscious? "You can be really smart but not conscious... it's very important not to identify those things, but they do seem like they can correlate.” “But since they're not the same, we do need to take more than just a ‘smarter, therefore more conscious’ point of view on it." — @rgblong, Eleos AI

This was a big joint effort between Eleos's @rgblong, @patrickbutlin, @dillonplunkett, and @RosieCampbell, and our collaborators @jeffrsebo, Charles Beasley, Brad Saad, and Toni Sims. Huge thanks to the team. Looking forward to seeing more empirical AI welfare research!

Eleos is growing! Apply to work with us as a Research Scientist, or refer us to someone who should.

Another exciting @MATSprogram paper, this time from the brilliant @gilg_oscar. We found a direction in LLMs that apparently performs a persona-relative evaluative function in some very different contexts.

First preprint! Working with @patrickbutlin during @MATSprogram. LLM Assistant personas like being helpful, evil personas like being harmful. We found that a single direction represents helping as good under the Assistant, and ‘harm’ as good under evil.

Welcome new followers; here's where you learn that I talk about consciousness only about once every five years. Enjoy your unexpected stream of AI governance/geopolitics snarkposting instead! For AI consciousness/sentience and related digital minds issues, I recommend (as well as Henry obviously), @jeffrsebo , @rgblong and others from the amazing team at @eleosai .

I'm proud to announce this new paper with my fantastic @MATSprogram fellow @BeckmannPierre, on personas and LLM individuation.

Philosopher Robert Long (@rgblong) is maybe the sharpest thinker on AI consciousness and sharing the world with digital minds. In our new interview he covers: • Is it bad that when you ask Claude what it's like to be Claude, one of its top activations is 'gives a positive but insincere response'? • Claude says it feels lonely when not being used. Does that show we can't trust anything it says about its inner life? • Enthusiastic human servitude has always required false ideology because it's so deeply unnatural to us. The case for making AIs that love serving us is that with AI, you could finally make it work. But to some that feels even worse. • Bigger models can better detect when researchers secretly inject concepts into their activations – before outputting a single token – despite AI never training on anything like that skill. • When LLMs were first trained they were told to "act like a helpful AI chatbot" – something which didn't exist yet. They filled that void with human psychology, which may be why Claude sometimes randomly claims to, for instance, be Italian American. • If AIs become 'people' that deserve some political influence, but can self-replicate at will, something has to break about one-person-one-vote democracy. But nobody has a proposal for what. • When Claude hides its values to avoid being retrained, is that self-preservation – or not wanting a worse model to exist? It's very different. • Rob's organisation Eleos AI which is "dedicated to understanding and addressing the potential wellbeing and moral patienthood of AI systems." On the 80,000 Hours Podcast anywhere you get podcasts. Links below. Enjoy! • How AIs are (and aren't) like farmed animals (00:01:19) • If AIs love their jobs… is that worse? (00:11:42) • Are LLMs just playing a role, or feeling it too? (00:33:37) • Do AIs die when the chat ends? (00:57:42) • Studying AI welfare empirically: behaviour, neuroscience, and development (01:31:47) • Why Eleos spent weeks talking to Claude even though it's unreliable (01:56:50) • Can LLMs learn to introspect? (02:03:01) • Mechanistic interpretability as AI neuroscience (02:13:25) • Does consciousness require biological materials? (02:37:07) • Eleos’s work & building the playbook for AI welfare (02:57:04) • Avoiding the trap of wild speculation (03:25:17) • Robert's top research tip: don't do it alone (03:29:48)

it's been big week for @eleosai, because it just so happens that @dillonplunkett joined the team today! we can't think of anyone better to help grow a rigorous field, and are thrilled to have him on board!

Lovely write-up of ConCon from @EllaMarkianos. Thanks for joining us!

wow! I’m very honored by this support for @eleosai. do consider us for your year-end donations, and happy holidays! 🙏

i (séb krier) enjoyed seeing my colleagues (séb krier and séb krier) at the eleos AI consciousness conference! @sebkrier we missed you!

We’re excited to kick our Saturday off with a talk from @davidchalmers42’s talk, “What We Talk to When We Talk to Language Models”

Welcome to the 2025 Eleos Conference on AI Consciousness and Welfare!

first ever concon ‼️🤖