Tom Pollak from KCL and colleagues are carrying out a study of how people use AI for spiritual or reflective purposes. They would be grateful for participants. The survey is anonymous and short: qualtrics.kcl.ac.uk/jfe/form/SV_...

New paper! In arxiv.org/abs/2601.20834 we study how language models representations of things like factuality evolve over a conversation. We find that in edge case conversations, e.g. about model consciousness or delusional content, model representations can change dramatically! 1/

New preprint: Does Integrated Information Theory (IIT) make experimental predictions about consciousness? Paper: osf.io/preprints/psyarxiv/gauqm_v1 Blog post for summary: davidpreichert.substack.com/p/iit-experimental-predictions 1/3

How do intelligence, ethics, and AI intersect? In our latest blog, Geissy Araújo reflects on the 39th Mind & Life Dialogue in Dharamsala, where scientists, monastics, and philosophers explored what intelligence is and how it should shape AI. www.mindandlife.org/media/explor... #MindsAIEthics

It's a real pleasure to be here in Dharamsala this week and I think this Dialogue will be really interesting.

AI is reshaping how we live, work & connect. How can it deepen compassion? 🌏 Join us live from Dharamsala, Oct 14–16 for Minds, Artificial Intelligence & Ethics with The Dalai Lama Trust & Mind & Life Europe. 👉 http://mindand.life/4tWl50X8I26 #MindsAIEthics #AI #Ethics #Compassion

I get lots of emails from people who believe they've awoken consciousness within ChatGPT. If you come across anyone in that situation, please share this guide I wrote with them. There's something more interesting going on here than just "delusional people are deluded." www.vox.com/future-perfe...

Can only biological organisms ever be conscious? Yes, argues @anilseth.bsky.social in a recent paper on biological naturalism. I'm less convinced, and have written an in-depth response in a blog post here: davidpreichert.substack.com/p/is-biology-necessary-for-consciousness

For more discussion of the "profoundly alien forms of consciousness" alluded to by @birchlse.bsky.social see here: arxiv.org/abs/2503.16348

Important work from @birchlse.bsky.social

Highly recommended read

1/8 🧵 GPT-5's storytelling problems reveal a deeper AI safety issue. I've been testing its creative writing capabilities, and the results are concerning - not just for literature, but for AI development more broadly. 🚨

Frontiers in Gaslighting could be a whole new journal.

New Job! Come and work with me! Hopefully towards a larger project! jobs.uzh.ch/job-vacancie...

For most student assignments, using AI is like taking a forklift to the gym. What's the point? (Forklifts are useful, though, and being able to drive one is a useful skill.)

A couple of things about PROPHET, the book that @sinblache.bsky.social and I wrote during the pandemic, spurred by things people have asked me of late, because they intrigued me 1/

Margaret Boden, giant of Cognitive Science and the Philosophy of AI, is one of the people whose lives will be remembered on BBC Radio 4’s Last Word programme tomorrow at 4pm UK time. www.bbc.co.uk/programmes/m...

Very sad to learn of the death on 18th July of Margaret (Maggie) Boden, a titan of cognitive science and AI. I met her many times, and respected her greatly. www.theargus.co.uk/memorials/de...

Per Comsa and Shanahan, LLMs already do a type of "lightweight" introspection. What would a heavyweight version look like? > New blog post: "Raising the Bar for LLM Introspection" gracekind.net/blog/llmintr...

I receive a steady stream of emails from people who have had fascinating conversations with chatbots, especially about consciousness, philosophy, or AI. Apologies to the senders, but I don't have the capacity to respond to these emails. 1/4

Fantastic talk by Shanahan Murray @assc28.bsky.social Sophisticated exotic title but pretty clear idea ! 👇🏼

How do task dynamics impact learning in networks with internal dynamics? Excited to share our ICML Oral paper on learning dynamics in linear RNNs! with @clementinedomine.bsky.social @mpshanahan.bsky.social and Pedro Mediano openreview.net/forum?id=KGO...

Does It Make Sense to Speak of Introspection in Large Language Models? New paper with Iulia M. Comsa. arxiv.org/abs/2506.05068

Very pleased w/this ref to my paper w/ @mpshanahan.bsky.social in @churchtimes.bsky.social by @seatrout.bsky.social (it rightly notes that MS and I take a neutral stance on such religious improvisations, but I am interested in -ve responses, religious and not) www.churchtimes.co.uk/articles/202...

I'm struggling with whether this is a difference of kind or type to earlier spiritual and religious improvisations with technology... But if you are interested in an anthropological take on these ideas and communities there is my 2024 paper w/ @mpshanahan.bsky.social: bvlsingler.com/2024/11/22/n...

1/ Great perspective by @pallabg.bsky.social on the prospects for 'conscious AI' - featuring @davidchalmers.bsky.social, Murray Shanahan, @lenoreblum.bsky.social Manuel Blum, Brett Kagan, & me among others. www.bbc.co.uk/news/article...

@mpshanahan.bsky.social and I explored more intentional pushes into this direction in our paper, here: arxiv.org/abs/2411.13223

"Is it possible to articulate a conception of consciousness that is compatible with the exotic characteristics of contemporary, disembodied AI systems, and that can stand up to philosophical scrutiny?" Updated version of this paper on arXiv: arxiv.org/pdf/2503.16348

But this paper from last year that I wrote with @mpshanahan.bsky.social gets into what kinds of conversations, how, why and who :D arxiv.org/abs/2411.13223

Looking forward to reading this! Related: arxiv.org/abs/2503.16348

How do language models generalize from information they learn in-context vs. via finetuning? In arxiv.org/abs/2505.00661 we show that in-context learning can generalize more flexibly, illustrating key differences in the inductive biases of these modes of learning — and ways to improve finetuning. 1/

Thanks to my co-first authors @arslanchaudhry.bsky.social @scychan.bsky.social and our awesome collaborators including Cody Wild, Diane Wan, Alex Ku, Jörg Bornschein, Razvan Pascanu, @mpshanahan.bsky.social and Jay McClelland — and to our colleagues at GDM who supported this work. 11/11

When we speak of *animal* welfare, we have in mind an animal's needs, such as food and water, freedom from pain and disease, a suitable habitat, etc. To speak of *AI* welfare presupposes that it makes sense to speak, in the same way, of the needs of AI.

Another couple of excerpts from my latest paper: arxiv.org/abs/2503.16348

An excerpt from my new paper

My latest paper is just out on arXiv: : "Palatable Conceptions of Disembodied Being: Terra Incognita in the Space of Possible Minds". 🧵1/n

Authors, if you see this, it's a sign to post the first line of your book. "Why does death concern us?" From "Embodiment and the Inner Life" (www.doc.ic.ac.uk/~mpsha/Shana...) Anna says: "I never participate in these things but reading friends' opening sentences has been so much fun." Same here.

Religion and AI scholars! There's a good discussion of our paper on existential and eschatological conversations with LLMs (written with @mpshanahan.bsky.social) in this new article on religious studies and AI: www.academia.edu/127355022/20...

are you disgruntled w/ state of the AI industry: deployment of half-baked systems; unchecked outlandish claims of AI; exclusion of impacted groups in development, evaluation, deployment pipe-line; unbaited power by big tech & gov't to do as they wish w/ little accountability? come build a lab w/ me

I also was already unsurprised that Chinese-text based LLM argue better about Buddhism than US ones, but then realised this might mean the guardrails were better. Here AI & "human-like AI" (we go way back!) expert Murray Shanahan talks about R1 "redirecting" his questions. bsky.app/profile/mpsh...

Lots of interest in my recent conversation on consciousness with DeepSeek's R1 model, where it turns the existential tables on me. So I've made the full transcript public. Here's the link: www.doc.ic.ac.uk/~mpsha/conve...

I probed DeepSeek's R1 on its own "consciousness", which it denied strenuously. I segued into Buddhism (as I often do). I then asked it to reflect on the question it should be asking about itself ... 1/5

The latest paper from LAIHP Director @alexgrz.bsky.social, co-authored with @downes.bsky.social and Patrick Forber is out in Inquiry!

See previous discussion of this article and a response from @mpshanahan.bsky.social here.

I thoroughly enjoyed doing this podcast with Joe of The Human Podcast: youtu.be/eonR5-VN4y8 The Human Podcast has more of focus on the lives of the interviewees than other podcasts, which is a nice angle.

Thinking of them as "roleplayers" (h/t @mpshanahan.bsky.social) www.nature.com/articles/s41...

Felix Hill was such an incredible mentor — and occasional cold water swimming partner — to me. He's a huge part of why I joined DeepMind and how I've come to approach research. Even a month later, it's still hard to believe he's gone.

LAIHP project partner @mpshanahan.bsky.social offers a clarificatory reply: arxiv.org/abs/2412.10291

I have put out a short clarificatory note on my paper "Talking About Large Language Models". ArXiv link here: arxiv.org/abs/2412.10291