iiinova — where your digital self keeps evolving

Jonathan Huston @jonathanhuston.bsky.social · 19/06/2026

"These four texts, Shelley's super-canon, instil in the Creature its notion of greatness based on classical ideals (Plutarch); a dramatic sense of creation and its moral challenges (Milton); a language of interiority and feeling (Goethe); plus a tragic sense of world history (Volney)."

aeon.co

000

Jonathan Huston @jonathanhuston.bsky.social · 05/06/2026

Sadly, not all writerly critiques of large language models are especially smart, coherent, or informed (to put it mildly). This one is.

theatlantic.com

No, Artificial Intelligence Is Not Conscious

Taken to its logical conclusion, this line of thinking is absurd—and damning.

000

Jonathan Huston @jonathanhuston.bsky.social · 16/05/2026

Maybe don’t read until you’ve read Transcription, but read Transcription. newleftreview.org/sidecar/post...

newleftreview.org

Lola Seaton, Beautiful Rules — Sidecar

Ben Lerner’s ‘Transcription’.

010

Jonathan Huston @jonathanhuston.bsky.social · 23/06/2025

Just as a minor point, you’re not a very close reader. My issue was with the narrowness of the writing clique as to who counts as a writer, not with narrowness of style. And anyone who claims that “anyone can publish” has spent very little time with writers who come from disadvantaged backgrounds.

110

Jonathan Huston @jonathanhuston.bsky.social · 22/06/2025

Spoken like a true conservative, bootstraps and all.

100

Jonathan Huston @jonathanhuston.bsky.social · 22/06/2025

If you prefer to read only those stories that people with sufficient unassisted writing skills are able to tell, you're free to do so. But it's not up to you to say "they don't need AI for that", since a lot of people do feel they would benefit from AI to tell their stories the way they want to.

200

Jonathan Huston @jonathanhuston.bsky.social · 22/06/2025

So if someone from a disadvantaged socioeconomic background has a story to tell that might be useful or educational or entertaining for the world to hear, you're going to require them to hone their craft and develop their skill as an artist first before you allow them to share their story?

210

Jonathan Huston @jonathanhuston.bsky.social · 22/06/2025

All fair points, but there are plenty of people who want (and should be able) to write who aren’t “writers” or “artists” in that narrow sense, and the Brevity article may be useful to them. The elitist comments have been along the lines of “if you need AI to write, you shouldn’t be writing”.

200

Jonathan Huston @jonathanhuston.bsky.social · 22/06/2025

Some of these comments are just dripping with elitism. Not everyone who has something to say benefits from the leisure, luxury, and MFAs to call themselves a “writer”. Telling them to just suck it up and “don’t write” is an indictment of how narrow the self-policing clique of writers has become.

300

Jonathan Huston @jonathanhuston.bsky.social · 21/06/2025

Yeah, there was no argument there, just self-righteous pontification without acknowledging her own hypocrisy. No, I’m not AI, but of course AI would say that. Yes, I rely on AI, but so do you by engaging with these posts.

000

Jonathan Huston @jonathanhuston.bsky.social · 21/06/2025

It’s also weird to me that her arguments against AI (not just on this thread) are highly moralistic, but she still advocates the use of AI-based technologies for fun, nothing she actually needs.

000

Jonathan Huston @jonathanhuston.bsky.social · 21/06/2025

I’m all in favor of using AI (and cars) wisely, but the premise here was that AI is completely useless and should be avoided entirely, while simultaneously praising a technology that relies on AI for its basic functionality. That’s not the same as saying: use AI, but be aware of the impact you have

000

Jonathan Huston @jonathanhuston.bsky.social · 21/06/2025

Oh, I get your point, but you don’t see any hypocrisy in saying 1) don’t use AI, it’s evil, while simultaneously saying 2) I use AI for the stuff I like? No straw man there that I can identify.

000

Jonathan Huston @jonathanhuston.bsky.social · 21/06/2025

Read “Attention is all you need” if you’re actually interested and don’t just want to throw facile insults around.

000

Jonathan Huston @jonathanhuston.bsky.social · 21/06/2025

I hate to beat a dead horse here, but you do realize there’s no “difference” between the two, just levels of sophistication? Generative AI is built on all the “old” AI technologies, especially deep learning, but also pretty much everything else you’d find in a pre-2017 AI textbook.

000

Jonathan Huston @jonathanhuston.bsky.social · 21/06/2025

Actually, she said all AI was useless and should be avoided, but I gave her the benefit of the doubt by assuming she meant generative AI.

000

Jonathan Huston @jonathanhuston.bsky.social · 21/06/2025

This was all in the context of the claim that AI is useless and evil, but still getting a kick out of using TikTok regularly. By using TikTok, you’re benefiting from LLMs and helping to train them. It’s like saying you shouldn’t drive a car because of the environment, but taking Uber everywhere.

100

Jonathan Huston @jonathanhuston.bsky.social · 21/06/2025

Moderation and recommendation rely heavily on LLMs now, even though they didn’t just a year or two ago. With the purported exception of Bluesky, those LLMs are trained on user data. And of course you already know that LLMs are just traditional deep learning on steroids.

000

Jonathan Huston @jonathanhuston.bsky.social · 21/06/2025

Nope, just years of experience working on this stuff.

100

Jonathan Huston @jonathanhuston.bsky.social · 21/06/2025

That’s probably largely true, but even scrolling generates user data.

110

Jonathan Huston @jonathanhuston.bsky.social · 21/06/2025

So you’re seriously claiming TikTok doesn’t use generative AI? Wow. You really don’t know how to look things up anymore, do you.

000

Jonathan Huston @jonathanhuston.bsky.social · 21/06/2025

TikTok/ByteDance uses LLMs extensively, not just “traditional” machine learning and symbolic AI, and is open about it. The research papers (and lawsuits) are freely available.

110

Jonathan Huston @jonathanhuston.bsky.social · 21/06/2025

TikTok uses LLMs extensively, not just “traditional” machine learning and symbolic AI, and is open about it. Why claim otherwise?

100

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

Because I think TikTok sucks? Um.

000

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

Yeah, poor choice of words on my part.

000

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

Again, “Wanda”, it’s so much easier to lob insults than to back up your stance. Who needs AI if you have people hiding behind a fake name who only know how to cuss? Nothing smart ass about it, just tired of the reflexive anti-AI hypocrisy, which is apparent from this thread. All slop, no facts.

100

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

She says generative AI is always bad, but TikTok is fun and she uses it all the time, even though TikTok relies on generative AI. What is there to misunderstand?

100

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

Probably in my mom’s analog attic :-). Somewhere between perceptrons and PyTorch…

000

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

True, I had email in 1992, Mosaic a couple years later. Set up GSA on the web in 1993.

000

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

But Eric, if you want to cite something showing that TikTok does not rely on generative AI, be my guest.

000

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

It’s been so easy for you to say “he doesn’t get the basics right” without getting the basics right. TikTok relies on generative AI, period, and you’re pretending it doesn’t, lobbing personal insults instead.

000

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

Yeah, but you quoted my response, so.

000

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

Generative AI is really just a sophisticated version of deep learning, as I’m sure you know. Bluesky’s algorithms are generative and leverage user data, they just don’t save user data for training purpose. Their wording is actually quite precise in this regard, to their credit.

010

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

It’s pretty easy to look up what kind of algorithms TikTok uses, but if you think it’s being a dick to point out blatant hypocrisy and privilege, so be it.

000

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

Nils Nilsson was my advisor, in case you’re actually interested.

000

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

I wrote one of the first neural networks to play Tetris and Othello (and beat humans) back in the 90s, using C and Assembly. Now mainly using Swift and Clojure. And of course AI tooling.

010

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

By following the links to the sources.

000

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

*sigh*. TikTok explicitly uses generative AI trained on user data, Bluesky uses deep learning that leverages user data to generate recommendations but doesn’t store the user data or train on it. Happy?

110

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

I guess you missed the part about “trained on user data”. Their ML doesn’t store user data, but it leverages it to make recommendations, otherwise it would be useless.

100

Jonathan Huston @jonathanhuston.bsky.social · 20/06/2025

Again, I’ve programmed AI since the 90s. I kind of know the difference. The big plus of Bluesky (unlike TikTok, the original topic of this thread) is that they don’t train on (i.e. store) user data, but the Discover algorithm would be useless if it didn’t use user data to generate recommendations.

100