Gwen's art

I don't do that, that's not my process. I always see people saying "well you're doing AI art the right way, it's only trained on stuff you own." I can never fall back on that. (And, honestly, even if I did use that process, I would be uncomfortable accepting such praise, since I think it is fundamentally based on a misunderstanding of the technology, and accepting that argument marginalizes AI artists who use a different process.)

I am not skilled as a traditional artist. I do photobashing and glitch art and digital collage. My process when I use AI is exactly the same as my process when I don't, it's just that ~50% of my source material now comes from an image model instead of from previously-existing images. In that sense I like to joke that it's actually less plagiaristic. When using AI, I source my non-AI material only from public domain image libraries out an abundance of caution.

I've thought about training a lora on my work. I think the results might be interesting - how would the technology respond to such a glitchy, noisy, messy, abstract style? But I wouldn't really be interested in posting the results here, because so much of what I consider my actual "art" is my collaging process. I'm not going to have the machine do the part that's fun for me.

So, like, I don't really have any moral high ground to stand on. I'm not disabled in a way that AI makes self-expression easier for me. (Although AI ought to have merit outside of being a disability tool.) I'm using AI based on huge internet scrapes. (Although a finetuned lora is also still based on huge internet scrapes.) Nothing about my workflow changes except that some AI image generation is added at the beginning, so it's not even ambiguous that I'm increasing my environmental impact. (Although it's still my opinion that the impact of AI image generation on the environment is negligible, and I derive enjoyment from it at least as much as I would from a graphics-intensive video game, so if gaming is morally permissible this is too.)