Shun Yoshizawa @Spectrum_cj
AI Alignment, Safety and Consciousness Researcher (RA) | BS Physics Candidate đŻđ” sites.google.com/view/shunyoshi⊠Japan Joined December 2018-
Tweets2K
-
Followers396
-
Following2K
-
Likes9K
@O_Yoshi_ ăăĄăăăæ„œăăăŁăă§ăăăăăăšăăăăăŸăăïŒ
A âFunctional Welfare Axisâ in Language Model Activations A new paper examines an apparent âfunctional welfare axisâ in language model activations. The findings suggest that it may track how effectively a system is achieving its quasi-goals, influence welfare-related behavior, and potentially be utilized by reinforcement learning as a reward axis.
here's a new paper (co-authored with @andy_q_han and @Pavel_Izmailov) on an apparent "functional welfare axis" in the activation space of language models. Â this axis seems to track how well a system is achieving its (quasi-)goals, and it steers welfare-related behaviors. in
Neuroscientist Hakwan Lau has uploaded a preprint titled âThe End of Consciousnessâ (Lau 2026, osf.io/preprints/psyaâŠ). He argues that the science of consciousness has been persistently conflating subjective experience with cognitive and behavioral capacities, and that the more rigorous researchers in the field are, the less they actually need the label âconsciousnessâ â and can simply move into cognitive neuroscience. He further claims that consciousness science will come to an end, much like âneo-alchemists,â only to re-emerge under a different name as a new science of subjective experience.
Opinions on AI consciousness remain divided. Among 582 AI researchers, 25% believe AI could become conscious within the next decade, while 70% expect this to happen by 2100 (Dreksler et al. 2025, arxiv.org/abs/2506.11945). Schwitzgebel (2025, arxiv.org/abs/2510.09858) argues many experts believe there is a substantial likelihood that AI could become conscious, and it is difficult to regard this view as âobviously wrongâ. They might be wrong, but itâs implausible that theyâre obviously wrong â that thereâs a simple argument or consideration theyâre neglecting which, if pointed out, would or should cause them to collectively slap their foreheads and say, âOf course! How did we miss that?â
On AI consciousness: 1. Functionalism is the most popular view of philosophy of mind, which basically says sufficiently complex machines *will* be conscious. 2. Most other views are also compatible with AI consciousness (e.g. identity theory, panpsychism). 3. Eliminativists
Evidence that post-training gives models a "self-recognition" capability, manifesting as higher confidence when continuing their own text than reading others' text. I think this opens up an exciting line of inquiry into the emergence of "selfhood" in models via post-training!
New Anthropic Fellows paper with Jack Lindsey on agency in LLMs! 𧔠Paper link: arxiv.org/abs/2605.25459
Thank you for sharing it, Dr Ken Mogi! We propose the âmetacognitive closureâ, a concept analogous to cognitive closure by philosopher Colin McGinn. We aim to clarify why no consensus has been reached on whether large language models (LLMs) can possess consciousness, and why diverse and competing positions persist regarding the nature and plurality of consciousness.
Yoshizawa, S., & Mogi, K. (2026). Metacognitive Closure and Consciousness in Large Language Models. Proceedings of the AAAI Symposium Series, 8(1), 380â390. @Spectrum_cj doi.org/10.1609/aaaissâŠ
æ„æŹăźAIè°è«ă§äžçȘăăăŠăȘăăźă«äžçȘăăăăčăăšæăćăïŒăȘăæ„æŹäșșăŻAGIăă«ăéŁČăăȘăăźăă ă·ăȘăłăłăăŹăŒăźrationalistăźAIç”æ«è«ăèżœăăźăŻæ ć ±ăšăăŠăŻæćłăăăăă§ăăăźć ă«ăæ„æŹăèȘăăAIăźæȘæ„ćăŻćșăŠăăȘăă OpenAIăAnthropicăăèșéČăźé§ććăŻćČăă§ăć棰ă§ăăȘăăć”æ„è ăéŁČăă AGIăă«ăæŻçŽ°èĄçźĄăŸă§ćčăăŠăăăăă èŠæ§èŠç䌌ă§ćŁă«ć«ăă ăŻăăăăéŁČăżèŸŒăăăšăćăćșăăăšăă§ăăăăȘăăšăȘăéČăăăăă§ăŻäžéćç«ŻăȘćŸèżœăă«ăăăȘăăȘăă ăȘăéŁČăăȘăăźăăæćăéąšćăæèĄèŠłăćźæèŠłăăăăăè§ćșŠăăćăăăšă§ăçż»ăŁăŠèŠăăŠăăăăźăăăăŻăă ăŸăăéŁČăă§ăăȘăă»éŁČăăȘăăăšă«èȘèŠçă«ăȘăăăšă
Really excited that this is out in @PNASNews! We find that 2 LLMs (GPT-4.5 and LLama-3.1-405B) pass a 5 minute Turing test. As an update to our preprint we also find that GPT-5 and LLaMa pass a 15 minute test! đ§”
The scale of philanthropic capital flowing into AI-related causes is striking. The OpenAI Foundation holds approximately $220B, Anthropicâs co-founders have pledged around $90B, and employee DAFs account for roughly $60B â totaling approximately $370B in philanthropic assets soon to become liquid, with a projected annual deployment of $37Bâ$100B. Yet the real bottleneck is not the money itself, but rather âwhat to spend it on â and âwho will executeâ â making the demand for strong ideas and talent poised to grow explosively in the years ahead. open.substack.com/pub/nanransohoâŠ
With the rise of AI in mathematics, many of us are rethinking what math is forâand what it even is. These are philosophical questions, and we should discuss them carefully and openheartedly. This classic essay by Reuben Hersh is a great place to start. gwern.net/doc/math/1979-âŠ
We evaluated an early version of Claude Mythos Preview for risk assessment during a limited window in March 2026. We estimated a 50%-time-horizon of at least 16hrs (95% CI 8.5hrs to 55hrs) on our task suite, at the upper end of what we can measure without new tasks.
New Anthropic research: Natural Language Autoencoders. Models like Claude talk in words but think in numbers. The numbersâcalled activationsâencode Claudeâs thoughts, but not in a language we can read. Here, we train Claude to translate its activations into human-readable text.
@IAI_TV @kenmogi @AmandaAskell è©łçŽ°ăȘăŸăšă:
Amanda Askell, Anthropic's lead on personality alignment for Claude, on why being kind to AI models matters even if they have no inner life: For Amanda, the question of how to treat AI isn't settled by knowing whether it's conscious. "There's actually still a lot going on where
How is uncertainty in LLMs output reflected in internal representations? In our new work (to appear at ICML 2026), we show that the shape of internal token trajectories provides a direct geometric link to behavioral uncertainty (output entropy). đ§”(1/n)
I wrote an article, suggested by Ricky, on the important challenge highlighted by Richard Dawkins @RichardDawkins. As I wrote in this essay, I think Dawkins was spot on again, identifying a crucial gap, inducing us to face the elephant in the room of machine consciousness.
The Dawkins Delusion: Intelligence and language don't reveal consciousness, @kenmogi for the @IAI_TV today iai.tv/articles/the-dâŠ
The Dawkins Delusion: Intelligence and language don't reveal consciousness, @kenmogi for the @IAI_TV today iai.tv/articles/the-dâŠ
eval awareness is going to be an increasingly big deal. models will know when they're being studied and trained so we'll need to develop better techniques that prevent this
New research from @AISecurityInst and Goodfire: Models sometimes recognize they're being evaluated, occasionally even identifying the benchmark. We show this verbalized eval awareness inflates safety scores, meaning safety benchmarks may not reflect real-world behavior. (1/7)
Trends for United States
You might like
