iiinova — where your digital self keeps evolving

Jim Buhler @jimbuhler.bsky.social · 08/10/2026

Except of course all those in academia who want to use AI to surveil the use of AIs by students (and others).

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

Yeah, I think the big problem with screens is that they distract other students along with the student who is using them. And that's true even when the student is doing something course related like looking up a term in wikipedia.

010

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

I have a hard time seeing the music AI stuff ever being financially viable. I think LLMs will survive on the strength of what they bring to coding. The LLMs ability to run grep and such also won't diminish with open weight models. So I think it will still be with us after the financial collapse.

120

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

Except isn't that what Fielding did? I mean we can wonder about what he thought he was doing with the original submission that was not disclosed, but the piece that appeared seems forthright in that respect. So we have his acknowledgment that he was initially embarrassed. But he takes that away.

110

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

The latter would involve feeding extensive notes and drafts, building context, iterating on prompts, very careful editing and rewriting, similar to collaborative writing and would involve disclosure. My sense is that Fielding's piece falls somewhere between the semblance of writing and this latter.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

I'm quite certain he did more than that. Still, it's interesting to think about where a line might be drawn, when what we might call vibe writing (simple prompts) becomes the semblance of writing (loose notes, general shape) and then again writing with an ethical claim on authorship.

230

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

And your work doing that (which I think is totally legitimate, btw) just makes the authorship situation murkier. That's the point I'm interested in—how AI (but also uncredited labor in general: see film music) upends accepted ideas of authorship. I think Fielding was trying to make that point.

000

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

Legally, there's no way to claim the LLM as a joint author.

000

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

I don't think anyone is denying it's a can of worms. It's very much a can of worms. That's why it requires thought rather than a snap judgment. That snap judgment is the way of the machine. We also don't know the extent of the prompting. Given the essay, I doubt very much it was a one-shot prompt.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

Fielding disclosed what he did, so here it's not a question of deception. At least legally authorship is not a measure of labor. Culturally I'm not even sure that authorship is understood as a simple labor test, as we permit all sorts of hands to participate without acknowledgement.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

The Nature article is revealing: if Fielding didn't author the article, it's hard to say who else did. It wasn't the LLM, because the LLM has no stake in it and can't answer for it. And Fielding is correct that the LLM would not have produced this article without his input.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

But is that sufficient? How is the responsibility for the work different from the responsibility of the work that Fielding took for the AI? And if you want to say Fielding exercised his control over a machine, what then is a camera, and editing deck, and projector if not an apparatus?

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

Does the director deserve "visual" authorship when you and I know the director may not even have been present at at least some of the filming, that there were other personnel involved in the visuals, it was edited by others, and so forth? I mean we can say that the credits carry that disclosure.

200

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

No, you marched in with an absolute general decree. The particulars refute that general decree.

200

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

The director is often given authorship of the film at a cultural level, though we both know the truth there. Screenwriting credit is so convoluted it requires a special process to disentangle. The question of legal authorship for a piece like this Nature article are still very much in dispute.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

I would say that a lack of disclosure is generally problematic. But I have difficulty getting upset that Brian Eno claims authorship for his generatively produced work; or that the guy claimed authorship over his Nature opinion piece.

000

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

This depends quite a lot on the circumstances. A celebrity memoire—I highly doubt anyone would bat an eye. A ghostwriter is practically expected. I think that's true of a lot of official correspondence too. And we are rarely outraged to learn that a politician's speech was written by someone else.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

You are just declaring by decree. Words are written in the name of all the time. Authorship is claimed for those words by the named author and culture accepts it, even in cases where ghostwriting is acknowledged. So in every meaningful sense the named author is understood to be the author.

200

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

That's not how it works legally (copyright) and often even culturally (the ghostwritten text, the political speech, or the complicated situation of something like a film). Another example: photography and how authorship works there either legally or culturally.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

It seems to vary a lot by discipline. Music students seem especially distrustful. But art students are much more embracing of it—at least the ones I met during a recent hiring committee I was on.

000

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

I'll be curious to see how well this works. So much of social media is now being filled with AI assisted or worse content, I wonder if students will feel any particular drive to make it fully their own? On the other side, I have noticed a strong anti-AI uptick on YouTube in the last few months.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

There are other technologies that are similar crutches to transcribing compared to AI in doing research. I have also been devising assignments that aren't exactly AI proof (some in fact require use of AI) but that can't be fully delegated to AI.

000

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

Yes, I use this analogy all the time too! You don't expect to go to the gym, watch your personal trainer do the workout, and expect to get the benefits!

000

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

In any case, all of this is on my mind because I'm teaching a course this fall that will at least in part be about AI and how to develop critical thinking around the technology and using it in research and writing. And, yes, one of the big things we'll cover is the dangers it poses.

210

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

That's why you need to teach students that! At least that's how I'm now thinking about it. I don't want to turn myself into a surveillance officer. I want students to determine their own ethic. A scarlet letter is a great way to increase anxiety but not a great way to develop a personal ethic.

200

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

I find it less helpful for literature reviews of a field I don't already know. It's ok for developing a basic research guide to loosely map a research area. Anything more than that I find I still have to do it mostly myself (though having the AI sort through my notes can be very helpful).

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

Verification is a skill you can teach, you know...

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

It's also an efficient way to run rabbit holes to see if they are productive. It's decent at sorting notes. It's true superpower is its knowledge of Unix commands.

000

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

LLMs are terrible at arithmetic, even if they are proving rather adept at coding and certain areas of mathematics.

110

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

I find it is decent as a sparring partner.

200

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

I mean, Google stopped working for me almost completely about 2 years ago, and that's one reason I turned to AI—because AI isn't terrible at that task, so long as you verify everything.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

Given that AI is getting more and more worked into our basic tools, I'm not sure I want to weaponize morality in this way. Is it ok to accept recommendations from autocomplete? what about copyediting? What about using AI to give you feedback on a draft? What about using it as a google replacement?

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

But how's that different from filtering AI output in that same way, by curation? I have lots of favorite bits of AI writing that I quite adore, even if I agree that the bulk is slop (like the bulk of human writing is slop).

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

I don't think we do understand that bad writing is bad. Or at least to the extent that we agree to something like that claim, we don't believe it is some objective universal standard. We might agree that everyone will mark a distinction but everyone will draw the line in a different place.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

I wouldn't agree with this claim. I don't think it's a new category. I would allow that we can call it the semblance of writing. But it's a semblance that is indiscernible from (at least some) human writing. You'll notice that even the best AI detectors have overlaps with some human writing.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

I think music generation is a far more complicated problem for a variety of reasons but at least in part because you are dealing with both composition and performance—or most music AI at least doesn't separate the two in a meaningful way. Text generation is a simpler task and has more training data.

200

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

I would say the problem is the slop—the most human thing about AI writing—not the AI. Slop is a machine issue only because the AI produces it at industrial scale. That's why we need to bring critical thinking to bear on the problem, not take the intellectual shortcut of taking the machine as proxy.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

AI music can now pass my indifference test (if I don't listen to it, I'm not immediately put off by it). Passing this test is recent but I've yet to hear a piece of AI music that gets beyond that. Text is more complicated in my experience. I think there's a lot of reasons why that's the case.

101

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

I will agree to this: I find that AI writing doesn't have a thought that moves through it. I'll add that I find that AI thinks the way that language thinks. Humans in general don't think the way that language thinks since they use language to think and human thought generally moves.

000

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

Those are pretty stark claims and I believe you are making the move of presence that poststructuralism long ago warned about. Can AI harm thinking? No doubt. Does it necessarily do so? Much more doubtful. The problem isn't the AI. It's the slop, and it's an all-too-human affliction.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

I agree that thinking is carried by writing and that writing bears the thought and its struggle to be expressed. I do not agree that writing only belongs to humans. To the extent that AI writing is cogent disproves that. I would say something like AI writing carries the thinking of language as such.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

That's a quality issue, and if you want to make a qualitative claim against his writing, that's completely fair. But make it a qualitative argument against the text, not an argument against the mode of production. Humans are fully capable of writing slop, and were fully capable of that before AI.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

You might look at the work of personal secretaries across history.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

He didn't say he didn't write his article. He claimed to stand behind and answer for the ideas expressed in it. That's why to the extent your critique has any force it bears on him: he, not his AI, is the author.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

Ghostwriting is a routine practice in many arenas. I would be surprised if a tenth of the opeds authored by our CEOs and politicians are actually written by them. They may be sketched out but the writing in almost all cases is executed by another.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

"at least are very hands on with them" — that's a huge concession in this context, since that was the very claim of the author you are condemning. Again we can talk about the ethics of writing in the name of another. What you are doing here is condemning him for disclosing.

110

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

We can talk ethics of all this writing in the name of, but we can't pretend it's a new issue, or that it wasn't common before AI. AI has simply made it more readily accessible. And that's a point worth pausing on and thinking about: why the panic now?

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

I mean, no. When a politician gives a speech written by a speechwriter, the politician is credited with that speech. When a president of a university has staff writers writing letters of thanks for a donation, the president is credited with that speech.

100

Jim Buhler @jimbuhler.bsky.social · 18/08/2026

The issue is more complicated than you are making it out to be. There is so much writing, long before AI, that has beeb done in the name of another. What makes delegating that task to the machine different from delegating it to the ghostwriter, the secretary, and so forth?

101