All art, to some degree, is farmed from public imagery. That isn't a great argument for AI art being theft. All major styles of visual art came from somewhere, and people copy them anyway. If you study art in college, art history is literally part of the curriculum and you spend time trying out different mediums and styles pioneered by other people.
Truly original art is rarer than 8' tall humans, but that doesn't stop the industry from existing at scale.
AI art is more akin to uploading a copy of someone's work and claiming it is my own that inspired work.
It would be one thing if the AI were actually trained like a human was and created unique artwork based on ideas given to it. The reality is it copy/pastes parts of images it found online.
> It would be one thing if the AI were actually trained like a human was and created unique artwork based on ideas given to it.
I am not sure you understand all this, because this is exactly what Stable Diffusion does. Give or take the “like a human does”, because we don’t understand how humans work.
> The reality is it copy/pastes parts of images it found online.
It does not. This is physically impossible; the pictures used during training are orders of magnitude larger than the model. It does not contain any description of those pictures used in training and you cannot get those pictures back.
What make you think human inspiration doesn't work the same way as image diffusers? To a degree we know we are compressing input features into our own latent space and conjuring them when producing our own works. The math may not be exactly the same, but the principles are eerily similar.
I don't work with them professionally, but I've done several small projects using them, and even written smaller models from scratch. I wouldn't call myself an expert, but I definitely know more than nothing.
Truly original art is rarer than 8' tall humans, but that doesn't stop the industry from existing at scale.