Originally written in mid-August 2025

Generative LLMs Have No Place in The Arts

The purpose of art, ultimately, is the expression of human interpretation. It is the act through which one conveys an idea to another. Read that again; capital-A Art is the act of the expression. This is why Art is so varied; it’s painting, it’s dancing, it’s acting, it’s directing, it’s woodworking, it’s many many things. This may seem abstract and philosophical on its face, but I think it’s actually quite intuitive.

Imagine a master woodworker building a chair. The materials they pick impact the sturdiness, rigidity, pliability, and appearance of the seat; the way they shape the wood, contour the seat and legs to make it more comfortable or support better posture; how tall they make the back of the chair, how heavy the final result is; all of these little choices add up to an expression of that woodworker’s ultimate design. While the final product - a chair - is mundane on its face, we would call this carpenter an “Artist” and we would be right to do so. We can examine their choices, sit in it, feel it, and in so doing experience what they wanted us to experience. Is this a desk chair? Part of a dining set? A throne? Did they choose the wood stain they did because they knew it would exist in a particular room and the light would catch it just so? And so on.

Conversely, imagine a chair factory with machines that churns out 10 identical seats every hour. We would not, intuitively, call this “Art” - but why? It is because Art is not the product, it is the human expression evident in the product. Nor would we typically think to call the people operating this factory “Artists”, because their expression is not evident in the chairs it produces; the chairs are simply built to spec, and if every worker in the factory were replaced they would still be identical.

There is still a capital-A Artist at work in this scenario - the engineer (or engineers) who set up the factory tooling. Perhaps they chose the particular wood they did because they knew it would remain sturdy through the rigors of warehousing; perhaps they implemented a clever design that allowed the chair to be easily assembled and disassembled, allowing it to ship or be fixed more easily. These are still human expressions; but we, mere chair-sitters, are not the audience - the factory workers are. The chairs themselves bear only the faintest vestigial remnants of these decisions, and so we would be right to not consider them Art.

I have often seen some philosophical quandaries - or, more often, impassioned defenses - of the use of generative LLM tools based on this idea that we humans are simply amalgamating past experiences and information when we create art, and the way these tools operate is in exactly that way. Who are we to claim what we make is art, and what these tools make is not?

I would respond to this with the above. “Art” is not a movie, it’s not a song, it’s not a painting, it’s not a game. “Art” is the intention behind those things that is evident or revealed within them. Generative LLMs fundamentally lack intention; as such, however technically impressive or intricate the results they produce, they will never be artists, nor will their output ever be Art.

There is one obvious response to this, which is, “Well, the human who prompts the LLM is the artist, and the prompt is their expression, so why does it matter?” And this is a fair question, one to which I’d ask you to recall the factory example above. The human prompting the LLM is not the engineer creating the tooling. The human prompting the LLM is the factory manager. The factory manager does just that; they ensure the requisite materials are available for chair production, they set the shift schedules, they ensure safety mandates are enforced. They do not, however, interact with the actual production of the chair. They do not make the choices in how to construct the tooling, how to design the chair for assembly, or shape the ergonomics of the seat. They are not an Artist - because they none of the choices involved in the Creation Of The Thing:tm:, obviously none of their choices are evident in the thing, and thus none of their expression is.

Using generative LLMs is much the same. The output is a production in which the human prompting it made none of the choices along the way. The implementation is the same regardless of which human wrote the prompt, so long as the seed is the same and so on.

The clear next progression is to bring up those humans that use these LLM tools in an iterative way - they edit outputs, stitch them together, reimagine them. Is this not capital-A Artistry?

Yes! I would say, rather obviously, that it is. And the more the human does, the more Art it is - definitionally. And thus we get to the final two points:

  1. Artistry is a spectrum. A programmer who prints the Mona Lisa in ASCII characters is an Artist, but so is Da Vinci, and we intuitively consider Da Vinci a “better” Artist and the Mona Lisa a more substantive piece of Art. There were rather more, and more complex, and more meaningful choices made along the way in one of those productions than the other.
  2. In spite of this, generative LLMs have no place in the Arts.

Why so? Not because they can’t be used, as part of a broader workflow, to produce Art, but because they aren’t. Humans are not using LLMs as an iterative tool in their toolkit that they occasionally use, occasionally partner with, and sometimes eschew at all. Humans are using LLMs to produce as much anatomically correct porn as they can masturbate to.

If you find a particularly high-minded individual, you might encounter one using LLMs to produce technically correct content that “looks right” or “sounds right”. The ultimate grail for all of these people I’ve talked to, their philosopher’s stone, is a LLM tool that produces suitable output in “one shot”; no editing, no prompt crafting, it just gets it right the first time.

The outputs from this quest are often technically accurate, intricately detailed, and what one might even call “impressive”. But they aren’t art. They’re intention-less; it’s just content, for its own sake.

This is what “AI slop” really means. It’s not that the pornographic image of Scarlett Johansson has 6 fingers on the left hand; it’s that it’s just content. No intention. No thought. No choices.

That is, just like chairs from the factory, something we would intuitively consider “not Art”.

And we’d be right.