What are the ethics of using machine learning (GANs and diffusion models) to change art styles of pictures?
That depends. Are you doing this just for your own enjoyment or trying to use that art for commercial purposes? Do you have the artist’s permission to modify and pass on their work?
If it’s just for you then go hog wild. If you’re planning to use the modified art commercially or post it online publicly you should get the artist’s permission first.
Strictly speaking on ethics, if it’s for some educational purposes then it’s fine. If it’s for the educational purposes of making a profit, then it’s not fine.
These models rely on pre existing art for their output, usually art which is taken without the artist’s consent, which is why a US District Court ruled that AI generated images cannot be copyrighted. If the AI’s output relies on pre existing art that has been taken without consent, then uploading it is largely considered to be art theft, and by changing the style that would also be considered modifying someone’s art without permission.
The way I see it is that it’s a bit like tracing. I’d be okay with someone tracing my art if it helps them practise, under the condition that they do not upload the traced piece or any parts that were traced from my drawings. If they uploaded something that has been traced, that would be art theft and I would be eligible to report it.
To just create your own versions, for your own use? No issues.
Do what you want 'cause a pirate is free
<3