And with the amount of corporate money in politics, and the current US administration, influencing legislation is a matter of writing a big check. I am pessimistic that we are on a good path.

(Chatham House rules means I can't say who, sadly.)

So the question of AI consciousness is driving legislation, and legal personhood is already being considered. While they haven't (to my knowledge) been linked yet, in law, they are absolutely being linked by legal scholars I've talked to in this space, e.g. at the Cambridge Digital Minds workshop.

Laws considering "legal personhood" are being considered, in Delaware for example; for now, those explicitly ignore the consciousness angle. Elsewhere, however, preemptive laws are being written to exclude personhood or sentience -- but they are older than Delaware's proposed laws I believe.

I don't "know for sure" that a tree or rock or set of inactive logic gates isn't conscious but it would be irresponsible for that level of uncertainty to support changing laws to take agency & responsibility from people (company CEOs) & give it to the spreadsheet. This is what is happening for AI.

As a scientist I generally agree with you. But as someone who cares that the headlines are screaming about AI consciousness (to get $$/eyeballs) & that policies are being proposed abt AI being "responsible" & having "agency" to enter into contracts (!!): ...

I don't feel unsure that my brain is conscious though. And you didn't answer my question 😀

And this isn't "intuition", unless you want to argue that a medical doctor uses "intuition" to treat disease. Are you suggesting that my literally being an expert in this field, leading a lab that does this, holding a doctorate, doesn’t at all license me to make statements based on that expertise?

Please read my preprint if you want to see why "it just is a different/simpler kind of C" doesn't solve anything. It just means there isn't one "natural kind" of consciousness & then we get into panpsychism & vibes again. If it is different, why call it C? Many, many others have argued similarly.

Which part of Claude do you feel unsure about? Which part is maybe conscious? The weight matrix? The snapshot of activation patterns (J-space or not) at timestep t? The update of those patterns at timestep t+1? I am curious: does your intuition suggest an Excel spreadsheet is minimally conscious?

I dunno man. One would hope. But I think at least some of them really believe it.

I've of course read the latest findings from Anthropic & elsewhere. But "threat representations" =/= phenomenal fear (Joe LeDoux). I'm an agnostic/weak illusionist so I don't think there is anything magical or ineffable about P-consciousness but I also don't see any evidence that Claude has C.

I know transformers are a different architecture. But AlphaFold has an attention mechanism. Vision-only transformers exist, with similar architectures to Claude. Are they conscious??

The same way I am confident that a really complicated, large, smart spreadsheet, or a deep neural net with recurrence & "internal representations" of what you show it, aren't conscious.

Please send any postdoc openings or faculty positions you know about our way -- US-based preferred, but he's willing to go elsewhere! 🙏 You and your group or department would be SO lucky to have him on your team.

w/ clarity of insight into theories of consciousness and wide background in sleep research. Background in EEG, fMRI, AI, ML -- he does it all. If I had more funding, I would keep him forever. (Until he wants to go be a professor I guess.)