AI and our appreciation of the arts and beauty in general

In My Humble Opinion
41

I really can’t make a judgement call about its fiction writing ability without seeing it write a cohesive story. It seems to have a knack for prose but that’s not the only thing needed to be a good writer. I know this because I know a guy who wrote beautiful prose and incredibly boring books. You would get excited about his work because the prose was so good, but it never went anywhere interesting and he seemed fundamentally incapable of learning.

I want to reiterate I never at any point denied the possibility that AI can make good “art,” only that I’ve never seen it. And that the storytelling I’ve seen from it is quite bad. But not inherently worse than what a human could theoretically create.

42

IMO, the difference is that a skilled artist using a paintbrush, camera, etc. has a clear idea of what results that tool will produce and how he can wield that tool to produce them. Whereas AI in its current form is still a “black box”: you can enter an incredibly detailed prompt specifying lighting effects, depth of field, etc….and still have only a vague idea of what the finished product will look like.

43

To @Dorjan 's point as well, there is also the application of AI tools (like Adobe for example) that uses AI in very specific and focused ways like removing or replacing backgrounds or objects from images or smoothing animation.

I think there’s also a utilitarian value in using AI to produce video or images. Like if I needed a short video of George Washington punching a dragon in the face for a third grade history class.

44

For what it’s worth, a 5000 word prose-worthy prompt is a terrible way to get what you want out of an AI image generator, even an LLM-driven one. Systems like ChatGPT or Gemini will internally crunch it down to maybe 200-250 words, disregarding much of your detail. Systems also get confused by that many tokens at once and lose coherence so a huge prompt asking for a red apple on a blue plate on a yellow tablecloth in a green painted room will result in it using those colors seemingly at random.

If you want specific complicated things that match your idea/vision rather than just whatever the AI coughs up, you’re getting into a lot of those other tools I mentioned earlier. A basic one is inpainting where you mask part of an image and just prompt for that location, adding an apple or turning a green apple red (or making it rotten, turning it into a baseball, etc). Then inpaint sketch where you draw on the image within the AI client to help guide it (make a red circle where you wanted an apple). You can take the image outside of the client and draw on it in another utility or add things in to photobash something closer then import it again via img2img and work off that. You can use utilities like ControlNet which then allows you to wireframe a person into a specific pose. If you are struggling with a specific item or style, you might seek out or train a LoRA to assist.

All of this is still AI image gen though. I suppose you might sketch or draw a little outside of it to assist the AI but that’s just giving nudges, not “drawing a picture”.

To be clear, most people don’t do all this stuff or anything like it. Most people just prompt and then prompt some more until they find something they can live with. However, people keep talking about super detailed prompts as though that’s the only option AI image generation gives you and that’s far from the case. It’s not even a good way to do it. But you can have very tight control if you want to apply the time and tools to use it.

I usually think of it as similar in some ways to staged photography. If you photograph a man on a balcony, you can pick the model, location, color of his suit, time of day, etc but you don’t decide on the size and texture of each brick, where the clouds are, the arrangement of the leaves in a tree, the shape of the skyline, etc. If the texture of his suit is important you can put him in silk instead of linen. If it’s not, you don’t worry about the fabric.

AI image gen can be the same way. The stuff you care about, the stuff that’s important to your idea, you can get exactly as you want it. The stuff that’s less important you can worry less about. You’re not “drawing” where you need to decide each millimeter of brick or location of each branch in a tree.

45

There’s very much a toupee fallacy going on with the general public. “I can spot AI art, it’s crap” when they spot low effort, low quality stuff and are unaware when they’re viewing the high quality stuff, which they may even enjoy unless someone tells them they’re supposed to hate it.

46

Fair correction. I’ve done very little image generation. I’m sure you’re right about which tools are most effective. That’s compatible and perhaps reinforces my point, in any case - people creating these things are putting in a lot of effort to make their creative vision manifest. But they would be dismissed by most people as “AI art” when it’s actually a creative human being using a set of tools that fall on the wrong side of what we’ve decided are the tools that can be used and still be considered art.

47

If it was just your comment, I would have let it slide (and did!) but I saw it a few times and figured it was worth correcting.

But, again, the viewer has no idea if I spent tens of hours, hours or minutes making an image. While I might know that what’s on the screen wasn’t just a lucky roulette spin, I have no real way of expressing that to the viewer. Such is a barrier to getting people to see it as “legitimate art”. Luckily, I (and people I know) just do it for our own entertainment or personal use and not for the accolades from the Art World :smiley:

48

I tried to cover this to a degree with my analogy early in the thread about the prolific artist cranking out what he knows is (competent) dreck and the starving artist pouring his soul into a piece for mediocre work. You can’t tell from the output.

And you may or may not be able to tell if someone using an LLM is using a complex set of tools (LoRas) etc to have extreme levels of control over the output that an LLM helps to generate.

My point is that there are lines to be drawn, that it’s a continuum and not a category. People that say that someone creating their own LoRAs, inpainting images, refining them, developing their own art style, etc. over long periods of time to create an image are not creating art, whereas humans more traditionally creating what they know is drek is art is arbitrary. It comes by the definitions you feel comfortable with. It avoids being thoughtful and challenging your own views. It’s a very simple, very popular view. It lacks nuance, it doesn’t measure the artistic or creative effort behind it, it’s elitist, and it’s generally not very well thought through.

That’s my whole point. It’s a continuum. Reasonable people can make their own judgments. It’s not categorical. It’s not “as soon as LLM even brushes up against something it’s automatically and definitionally not art”

49

That’s an interesting point. There’s a broad spectrum of how much AI an artist has put into a piece.

I’ve been reading some articles about books getting published and then pulled because it was found they used AI. I’ve been thinking: used it how? Plot brainstorming? Outlining? Thesaurus? Full draft generation where they basically just signed their name? Post-draft line editing? I feel like we’re losing perspective and it’s kind of silly. I don’t love AI by any means, I’m a bit of an AI curmudgeon, but even I can see the qualitative difference in those things. And I’ll also add that a lot of major writing programs have AI integration pretty much without consent.

There’s a very popular writing program far predating AI
where you can put your MS through analysis and make changes the program recommends along like twenty different dimensions. No one ever thought twice about using that. I stopped using it because whenever I “fixed” my manuscript it no longer looked like anything I’d write. Once I got frustrated and cranked my favorite Stephen King short story through its analyzer. The program pointed out all kinds of problems with it, which told me about all I needed to know.

Is that cheating? If that’s cheating, how is it any different than using a line editor? Is it wrong to use AI as a line editor just because it’s a machine and not a person?

See what I mean?

50

One thing I’ve been noticing lately is a number of people who have, apparently, decided that if “AI art = bad art,” then “bad art = AI art.” I saw a post on Reddit for someone’s self-published D&D supplement, and the cover art was very amateurish. Proportions weren’t right, the line work was sloppy, and the colors uneven. And a bunch of people were calling it out as “AI slop,” when those are not the ways that AI is bad at producing art. When someone posts an “Is this AI?” questions, people zero in on the hands, because AI has famously struggled with that - but so do humans. Hands are hard to draw! Just a couple days ago, a post that was full of grammatical errors and spelling mistakes had people were saying it was AI generated text, when AI text generators are vastly better at grammar and spelling than that post.

51

I should retract part of my OP. I actually do see a lot of creative work done by creative people on AI. I even enjoy the instrumentals I hear. It’s is the voice I don’t like and in most cases stories if I know it to be AI, if I don’t know I am fine with it.

52

The NYT article I mentioned, gift linked.

https://www.nytimes.com/2026/03/19/books/ai-fiction-shy-girl.html?unlocked_article_code=1.UlA.AvHm.cL2GvCAbswBB&smid=nytcore-android-share

“AI is writing fiction. Publishers are unprepared.”

Essentially publishers don’t want to set a hard and fast rule against AI or draw any lines, which creates ambiguity for authors and readers. In the case discussed here, it appears the book was pulled from publishing after an acquaintance of the writer used AI to edit her manuscript.

The article points out that anything written by AI isn’t copyrighted so publishers are hesitant to use it, but they won’t outright prohibit it either.

Which raises the question of how much or to what extent something is touched by AI loses its copyright.

53

If an AI programmer uses AI to create a painting, and I then push the same buttons and provide the same source material, am I then the creator of that painting? Am I the artist?

54

I’d say that you’re the creator of the image but perhaps not the artist.

If you follow directions for setting Lite Brite pegs or a cross stitch pattern, you created the resulting display but you didn’t put your own creative thought or vision into it. I’d say the same applies if you follow someone else’s Prompt+Model+Seed+Settings in an AI client.

Now, if you both did it independently of one another then I guess you could both be the artist though that’s exceedingly unlikely especially when it comes to seeds.

55

So I just signed up for Midjourney and it’s incredible and it reminded me of this thread. I wanted you guys to take a look at some of the things people are creating with midjourney and see if you think it’s “AI slop”

You have to sign up to be able to view the feeds, I think. You don’t have to pay and you can use a google o-auth login, so it’s like a 2 click process.

And using the tool is INCREDIBLE. far better than any other AI image generator I’ve ever used. This little 150 man company with no venture capital (just user subscriptions) has created the best image generator on Earth. Better than google, better than openAI. It’s beyond impressive.

also, I challenge you to look my little cheesecake kitten in the face and tell him that he’s slop.

56

I have to agree with you, and I am rapidly beginning to accept the creative side of AI. I see it all around me and it gives those who use it a kind of persona builder. Our art reflects who we are and what we like.

57

I appreciate you being open minded. I don’t like how many people instantly jumped on the “AI is terrible, AI is slop” train. There are legitimate downsides to how AI might shape society, but there’s also a lot of really cool shit you can do with it. To a degree, there’s garbage in garbage out - if you generate thoughtful prompts and conversations, you get high level engagement. This is true beyond image generators - when I try to approach an intellectual debate with an AI, I try to approach it like I would in great debates on the SDMB - state my premises, try not to use fallacies, clearly make my arguments - and return it gives me extremely high level discussion. I think a lot of users give it very basic, undisciplined, and uncreative prompts and then say “eh this isn’t that great” when they get garbage out.

If anyone is interested in creating images and short videos (about 5 seconds), I highly, highly recommend throwing $10 at the midjourney base subscription to try it out. You get plenty of gpu time to try it out with that. I started there and upgraded to the $30 subscription because this is one of the most fun things I’ve ever done with a piece of technology and it’s absolutely worth it.

58

I think anyone can see it if they have a link, you just can’t extend the feed (“See More”). https://www.midjourney.com/explore?tab=hot is the link to the current upvoted images (which I think are a better example than the video clips)

I do most of my image gen via local setup but still keep a $10 sub to Midjourney for knocking around on Discord, ideas and other stuff. My impression is that the images are generally better than most other services though it suffers from prompt coherence without a giant LLM backing it up and currently text generation kind of sucks (though v8 will be better and is due out soon). If your goal is to see if Midjourney can prompt a tongue-twister correctly vs ChatGPT or Gemini, it won’t. But what it does create will typically look a fair bit nicer.

People who kneejerk call anything generated by AI “slop” typically aren’t worth engaging with on the topic. That said, most of what people see made via AI is low effort one-shot memes and joke images and the image quality from earlier models often wasn’t great. Add to that, people just not bothering to give a visual go over for flaws and trying to fix them and you do have a lot of trashy images. I get and appreciate that. But it’s certainly not the best AI is capable of, just like you wouldn’t judge photography based on a bunch of crappy old iPhone images or 35mm disposable cardboard cameras collected from a wedding reception.

59

My publisher outright prohibits the use of AI. But publishers are no longer gatekeepers. Another Times article describes a romance writer who self-published 200 books last year. She claims she carefully edits each one, but that seems unlikely. No publisher involved. She doesn’t sell a lot of most of them, but together they add up.

Anyhow, maybe the discussion should be what is art. The definition seems to change over time. Could AI easily generate soup cans or images of Marilyn Monroe? Do we call that art because a person did it?
It reminds me of the early days of computer chess, where skeptics thought chess was a human activity people could never do. They’d laugh at obvious mistakes in the early programs. Then the programs beat the skeptics. Then the programs beat the grand masters. When AI generated art of fiction gets better than most humans can do, will we call it real?

60

I heard this from somewhere on the internet so I can’t take credit for it.

But the theory was that the purpose of art is to see and experience the world through another person’s eyes. It could be a story told through their point of view or a fictional story presenting a framework for viewing the world. But the point is to enable people to connect on ideas and perspectives, even if they don’t agree with them or even find them distasteful.

So it sort of begs the question what is the point of AI generated art? Even if it’s very pretty or technically proficient, it’s not representing anything other than algorithmically generated noise,