1130
Consequences
(pawb.social)
Source (Bluesky)
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 04 Sep 2025
1130 points (100.0% liked)
Fuck AI
3976 readers
736 users here now
"We did it, Patrick! We made a technological breakthrough!"
A place for all those who loathe AI to discuss things, post articles, and ridicule the AI hype. Proud supporter of working people. And proud booer of SXSW 2024.
founded 2 years ago
MODERATORS
It's also important to remember these models are trained by sampling (imitating aspects of) images they don't have the rights to use directly. I think it's justified being angry about someone using your work -insignificantly mashed together with millions of other people's work- without your permission, even if it's to extend a background by 10 pixels lol
So is basically every human artist. Basically any artist out there has seen tons of other art prior and draws on that observed corpus to influence their own output. If I commissioned you to draw something you didn't know what was, you'd go look up other depictions of that thing to get a basis for what you should be aiming at.
The way AI does it is similar, except that it's looked at way more examples than you but also doesn't have an understanding of what those things actually are beyond the examples themselves. That last bit is why it used to have so many problems with hands, and still often has problems with writing in the background or desk/table legs.
We can actually look at a hand, and understand it, logically thinking about the composition and style to work with. AI can only copy paste the difference of pixels' colors on digital images whose metadata happens to contains the word 'hand'. No matter how many 'examples' have been scraped, it can't actually interpret them the same way we do.
If some alien species asked you to draw part of it's anatomy that can move into a wide array of configurations, but you are required to do so based only on pictures the aliens sent you that they tell you shows that part among other things, would you do better?
Like, what you said is specifically why it's bad at hands and table legs and the like - they can appear in many different ways and it's only reference point for them is pictures of them it's seen. You understand hands and think logically about them mostly because you have a not just wider but deeper set of experiences to work from. Even then, 4 fingered hands have been common in cartoons because even having hands, being surrounded by other beings with hands and in a culture that makes heavy use of hands a lot of artists have trouble doing them quite right.
Yes, I would do better. I would take a look at the pictures, and think about the angles / geometry, the reason of differences between the pictures, and being able to count sure helps. If they were to show me pictures in a vastly different style, I would make assumptions, like it is a different representation of the same concept. I would not just mash them together based on color values.
I get what you're coming from, but the only reason these models seem to be able to get stuff done, is the insane amount of training data and iterations.
Enjoying this discussion, by the way! It's fun to think about.
Not all them. Some are trained on pure public domain data (though admittedly most folks running locally are probably using Flux or Stable Diffusion out of convenience).
And IMO that’s less of an issue if money isn’t changing hands. If the model is free, and the “art” is free, that’s a transformative work and fair use.
It’s like publishing a fanfic based on a copyrighted body. But try to sell the fic (or sell a service to facilitate such a thing), and that’s a whole different duck.
50 Shades of Grey be like...
A surprising number of people don't know that it's essentially Twilight fanfic with the names changed to protect the author from being sued.
Yeah, that's an interesting case.
I guess there was no incentive for Stephenie Meyers and E. L. James (and their movie adaptation money banks, Lionsgate and Universal) to sue. But apparently it was brought up in some kind of lawsuit over an actual pornographic adaptation:
IMO, training software on the corpus of human art without payment or attribution is not good for society and art in general, but, humans who create non-abstract are trained and honestly create in a strikingly similar way. The person hired to make an art piece of Catherine the Great doesn't disclose that he looked at Alexander Roslin's painting of her and is greatly copying the look and feel for the face or the Google search they used to find options for 1700's royal clothing. The big difference in process between AI and an artist with reference art is the removal of the human element, and that's super important.
But instead, we focus on how it was trained, when we train much the same way, or we call it all slop regardless of the actual quality, instead of calling out the real problem, the one problem that we can do something about, it's taking a living away from humans.
Slop machines have more in common with tracing than with learning.
https://www.youtube.com/watch?v=iv-5mZ_9CPY
You can hate the output, you can hate copying the IP, you can hate the people involved, but the process is nothing like tracing. It's closer to black fucking magic :)
It's not magic.
You're in a cult.
It's closer to black magic than tracing. It's really closer to sculpting. You take a machine, show it a few million pictures of dogs to teach it what dogs look like, then hand it a big block of white noise and ask it to show you a dog. It then carves away all the parts it doesn't think look like dog, repeats until the result doesn't change much and shows you the dog it essentially hallucinated out of the noise.
funny, I didn't feel like a kettle this morning.
anyway have a good one
for 1700s* royal clothing