iiinova — where your digital self keeps evolving

Will Fleisher @willfleisher.bsky.social · 18h

Yeah, I didn’t have any doubts. But my own proclivities are to be pedantic af, so. Here we are.

Will Fleisher @willfleisher.bsky.social · 18h

I feel like this is a pretty standard view. It’s better to be someone who convinces the predictor you will one box, and then to two box. But that’s not *really* part of the example. It’s like saying “I would derail the trolley”.

160

Will Fleisher @willfleisher.bsky.social · 19h

Also, I love Jemisin. And I would like to see if people make suggestions.

120

Will Fleisher @willfleisher.bsky.social · 19h

If you’re asking this question, I’m sure you’ve read them already, but in case there are other people about: Octavia Butler’s three main series all engage with both things. Also, some of my all time favorite books.

140

Will Fleisher @willfleisher.bsky.social · 14/10/2026

Something something McCarthyism, so really both sides… etc.

000

Reposted by Will Fleisher

Greenb3ar @greenb3ar.bsky.social · 13/10/2026

Okay the best political ad of the year is here from LA-DSA

124119774361

Will Fleisher @willfleisher.bsky.social · 13/10/2026

And then all the other physicists and philosophers of physics had to spend all their time talking about how the perpetual motion machine is still very much not going to happen.

000

Will Fleisher @willfleisher.bsky.social · 13/10/2026

I don’t have ADHD, to my knowledge, but I feel this…

010

Reposted by Will Fleisher

noam @noamchompers.bsky.social · 10/10/2026

someone should tell my man that people deprived of USAID believe that they are experiencing pain

1383

Will Fleisher @willfleisher.bsky.social · 05/10/2026

Apologies for the long thread... this is why I resisted Noam's original request to elaborate. I knew it would take me awhile to write it out. Perhaps you have an easy response, or perhaps its better to write about this after the paper is already published.

010

Will Fleisher @willfleisher.bsky.social · 05/10/2026

... even after making the distinction between core realizers and total realizers. That, again, is not to discount the value of that distinction. It will probably help a lot in making further progress. But I don't think that, as of yet, it defuses the cheap consciousness worry.

200

Will Fleisher @willfleisher.bsky.social · 05/10/2026

...(e.g., a global workspace), but which clearly aren't conscious, this provides disconfirming evidence for the package of GWT combined with the relevant auxiliary hypotheses about what the background conditions are. So, I think the cheap consciousness problem is still a problem for those theories..

100

Will Fleisher @willfleisher.bsky.social · 05/10/2026

Second, I think leading computational functionalist theories do have commitments about the enabling conditions. They tend to be commitments that the systems in question must be a certain sort of computer. If we can build sophisticated *computers* that have the right core realizers per the theory...

100

Will Fleisher @willfleisher.bsky.social · 05/10/2026

... I gain evidence about it by testing it in conditions where the enabling conditions are met. (Or if I just know enough about the world to know which enabling conditions are relevant). I don't see a plausible route to learning the theory without also learning about the enabling conditions.

200

Will Fleisher @willfleisher.bsky.social · 05/10/2026

...requires testing them along with auxiliary hypotheses, in this case specifically about the enabling conditions. This is for traditional confirmational holism (Duhem-Quine) reasons. Take your own helpful case about braking systems. The only way I could know a system is fit for braking is if...

110

Will Fleisher @willfleisher.bsky.social · 05/10/2026

... don't know whether an AI has conscious states. But those scenarios involve learning the true theory by oracle or divine providence. In the nearby worlds where we actually learn the correct theory, we will also almost certainly know the enabling conditions. This is because testing our theories...

100

Will Fleisher @willfleisher.bsky.social · 05/10/2026

First worry is about this claim: "that even if we had the correct theory of consciousness [...] we still wouldn’t know whether artificial systems have conscious mental states". I think this claim isn't true. It *is true* that there are imaginable scenarios where we know the right theory, and...

100

Will Fleisher @willfleisher.bsky.social · 05/10/2026

Well, as I said earlier, I think the paper is good. I'm not sure responding to these objections in the paper will make it better. Rather, it might be that these objections are better raised in a future paper, to which you might respond. (Or perhaps not; I've been wrong before). But, a short version:

100

Reposted by Will Fleisher

Annalee @flowerhorne.com · 05/10/2026

Begging news media to stop making themselves unwitting NPCs in a weird cult's alternate reality game You do not need to breathlessly repeat claims that LLMs will end the world or create utopia with handwavy details or abstract philosophical arguments Ask follow-up questions or hire people who can

2690159

Will Fleisher @willfleisher.bsky.social · 05/10/2026

How about just a direct democracy of people who actually value knowledge and expertise and voluntarily listen to it.

110

Will Fleisher @willfleisher.bsky.social · 03/10/2026

Yeah, definitely obscure as written, sorry. But I think a bsky thread probably isn’t the easiest way to convey it. I will work on it.

130

Will Fleisher @willfleisher.bsky.social · 03/10/2026

Still actually a better theory than contemporary theories of consciousness.

010

Will Fleisher @willfleisher.bsky.social · 03/10/2026

… the true theory we wouldn’t be able to tell which systems were conscious is not well supported. The main issues, in both cases, are methodological, and have to do with confirmational holism. Obviously I would have to say more to back this up. But maybe I will, I’m working on related stuff.

120

Will Fleisher @willfleisher.bsky.social · 03/10/2026

This paper is good. The point about core realizers is really important and not often recognized in the literature. However, I think his arguments don’t ultimately go through here. Cheap consciousness is still a damning problem for these views. And I think the claim that even if we had….

200

Will Fleisher @willfleisher.bsky.social · 02/10/2026

Oh no! I don’t mind being wrong. But naive? According to social psychologists? How embarrassing for me.

010

Will Fleisher @willfleisher.bsky.social · 02/10/2026

I mean, as long as our affective associations are tuned up so as to be positive toward epistemically good features, I don’t see any problem. I have positive affect about elegant and simple theories that are still empirically adequate. As long as they are truth apt, I’m in good shape.

110

Will Fleisher @willfleisher.bsky.social · 01/10/2026

The models might well characterize what we do, as a real pattern, even if most of us aren't explicitly following them. IBE describes what many of us think we are doing, but you can reframe it without loss using a Bayesian model (I know that's somewhat controversial, but also I think it's true).

120

Will Fleisher @willfleisher.bsky.social · 01/10/2026

Eh, just slap the IBE part into the likelihood part of your favorite model of confirmation, help yourself to some of the evidence as evidence in this model, and now you don't have to be committed to IBE as a separate kind of argument.

120

Will Fleisher @willfleisher.bsky.social · 01/10/2026

I think you all might be overly restrictive about what counts as an argument. "Look how nice things look from this view" is a great argument, assuming you cash out "looks nice" properly. If "looks nice" means *is empirically adequate and has these awesome theoretical virtues*, it's a great argument.

180

Will Fleisher @willfleisher.bsky.social · 27/09/2026

I take it there isn’t really a lack of demand among foreign students, in the sense that if the US stopped having terrible xenophobic policies we could have plenty of students.

000

Reposted by Will Fleisher

Carl Zimmer @carlzimmer.com · 27/09/2026

Four days ago, a scientist was surprised by the news that Anthropic discovered new virus genes for making DNA. He says he’s been studying them for years—and feeding his data to Anthropic’s AI as part of his own research. Coincidence? Here’s my story. nyti.ms/3VSWayc

nyti.ms

Did Anthropic’s A.I. Really Make a Scientific Discovery on Its Own? (Gift Article)

An expert at the University of Copenhagen said his team had been sharing its research with the company’s A.I. model, Claude, and that its new finding matched their work.

882124874

Will Fleisher @willfleisher.bsky.social · 26/09/2026

Granny and Aunt Bertha would be proud.

010

Will Fleisher @willfleisher.bsky.social · 25/09/2026

People worry that they will be judged if they admit to using AI for writing. They are correct to worry about this. I will in fact judge you if you use AI for your writing or if you advocate for more permissive norms for its use. I am judging you right now.

040

Will Fleisher @willfleisher.bsky.social · 25/09/2026

We can push for changes to these things via regulation. (Or my preferred solution of nationalization).

000

Will Fleisher @willfleisher.bsky.social · 25/09/2026

The particular proposal people are talking about isn’t a good idea. But generally, people know the difference between figurative uses of mental terms and literal ones. The problem is that the AI industry PR and their product design are trying to confuse the distinction.

100

Will Fleisher @willfleisher.bsky.social · 24/09/2026

At the HAA, they throw all the Halloween candy away.

030

Reposted by Will Fleisher

Philip Sink | Rorty Vindicator @thrillhouse1701.bsky.social · 23/09/2026

'Follows from a more general principle' is a timeless banger and if someone doesn't get it that's their problem.

0111

Will Fleisher @willfleisher.bsky.social · 23/09/2026

Or in the parlance of our times “This, but unironically”.

020

Will Fleisher @willfleisher.bsky.social · 23/09/2026

I didn’t know I needed this image, but I did. Fantastic.

131

Will Fleisher @willfleisher.bsky.social · 23/09/2026

Basically how I conduct every class.

000