I've gotten arguments that it's theft, because technically the AI is utilizing other artist's work as resources for the images it produces. I've pointed out that that's more like copying another artist's style than theft, which real artists do all the time, but it's apparently different when a computer algorithm does it?

Look, I understand people's fears that AI image generation is going to put regular artists out of work, I just don't agree with them. Did photography put painters out of work? Did the printing press stop the use of writing utensils? Did cinema cause theatre to go extinct?

No. People need to calm down and stop freaking out about technology moving forward. You're not going to stop it; so you might as well learn to live with it. If history is a reliable teacher, it really won't be that bad.

[–] Monstrosity@lemm.ee 11 points 1 month ago* (last edited 1 month ago)

Well said.

I'd like to add that the biggest problem, imo, is the closed source nature of the models. Corporations who used our collective knowledge, without permission, to create AI to sell back to us is unethical at best. All AI models should be open source for public access, sort of like libraries. Corpos are thrilled we're fighting about copyright pennies instead, I'm sure.

[–] EldritchFeminity@lemmy.blahaj.zone -3 points 1 month ago (2 children)

Except it isn't copying a style. It's taking the actual images and turning them into statistical arrays and then combining them into an algorithmic output based on your prompt. It's basically a pixel by pixel collage of thousands of pictures. Copying a style implies an understanding of the artistic intent behind that style. The why and how the artist does what they do. Image generators can do that exactly as well as the Gaussian Blur tool can.

The difference between the two is that you can understand why an artist made a line and copy that intent, but you'll never make exactly the same line. You're not copying and pasting that one line into your own work, while that's exactly what the generator is doing. It just doesn't look like it because it's buried under hundreds of other lines taken from hundreds of other images (sometimes - sometimes it just gives you straight-up Darth Vader in the image).

[–] Honytawk@lemmy.zip 4 points 1 month ago* (last edited 1 month ago)

It’s taking the actual images and turning them into statistical arrays and then combining them into an algorithmic output based on your prompt.

So looking at images to make a generalised understanding of them, and then reproduce based upon additional information isn't exactly what our brain does to copy someones style?

You are arguing against your own point here. You don't need to "understand the artistic intent" to copy. Most artists don't.

[–] desktop_user@lemmy.blahaj.zone 2 points 1 month ago (1 children)

and just about any artist can draw Darth Vader as well, almost all non "ethics" or intent based argument can be applied to artists or sufficiently convoluted machine models.

[–] EldritchFeminity@lemmy.blahaj.zone 2 points 1 month ago (1 children)

But just about any artist isn't reproducing a still from The Mandalorian in the middle of a picture like right-clicking and hitting "save as" on a picture you got from a Google search. Which these generators have done multiple times. A "sufficiently convoluted machine model" would be a senient machine. At the level required for what you're talking about, we're getting into the philosophical area of what it means to be a sentient being, which is so far removed from these generators as to be irrelevant to the point. And at that point, you're not creating anything anyway. You've hired a machine to create for you.

These models are tools that use an algorithm to collage pre-existing works into a derivative work. They can not create. If you tell a generator to draw a cat, but it hasn't any pictures of cats in its data set, you won't get anything. If you feed AI images back into these generators, they quickly degrade into garbage. Because they don't have a concept of anything. They don't understand color theory or two point perspective or anything. They simply are programmed to output their collection of vectorized arrays in an algorithmic format based upon certain keywords.

[–] Honytawk@lemmy.zip 5 points 1 month ago (1 children)

Why are you attaching all these convoluted characteristics to art? Is it because you are otherwise unable to claim computer art isn't art?

Art does not need to have intent. It doesn't need philosophy. It doesn't need to be made by a sentient being. It doesn't need to be 100% original, because no art ever is. So what if a computer created it?

If you encounter an artist who never saw a cat, they would also not be able to paint it. Just look at these medieval depictions of lions where it is clear the artist never saw one.

[–] FlyingSquid@lemmy.world 2 points 1 month ago

Also, Dürer's Rhino.

But he tried. That's the important thing.