Edit: First and foremost, I'm sorry you felt attacked. That is never OK. I need to step back from posting anything until I can be a decent human being. I regret not doing that before I posted.

----- People claiming you can copyright a canvas that is just a color (when you can't, you can only copyright the art installation/display). People claiming that you can copyright the alphabet (when you can't). It's just frustrating that HN wants to keep having this discussion but with ZERO basis in actual copyright law, and people making factually inaccurate claims as if they understand it. I had the same issue on a criminal law post. Someone posted completely factually inaccurate information regarding title 18 statutes, and HN blocking me from responding in a timely manner while people were reading the post. This just isn't a forum for informed discussion I guess, but peoples gut feelings and what they THINK copyright law is. Opinions are great, and needed, but established law is a real thing and should be part of a discussion that at it's core is about copyright.

Look at the inputs and outputs of generative AI art - specifically the ones shown in the post and others. The disconnect here is not our insights on copyright law (At least not for generative art. Maybe for copilot, code, and GPT3, but not for the art.) Hollie is a Character Artist. A Character has a combination of appearance, emotional radiance, mood, behavior, that have to fit in to create a unique persona and trigger an emotional response to that character from viewers. Stable Diffusion does not take or copy her characters - it mimics her drawing style in colors, hues (it reads what pixels are proximate to what other pixels and tries to apply that proximity to similar colors, line patterns in the future). When you create a stable diffusion image, you set a parameter of how much randomness to apply and how much variety, and what else to mix in. So any content it's trained on becomes a soup of colors and lines and hues and saturation. So if Hollie draws with de-saturated colors, and favors pinks and greens, an image produced with her name will probably have that, but not more than a soup mixed with objects searched by the words of the generator human. Code generation is different as a code has to appear in some form to work properly and that happens with a lot less variance, so generators there would likely remix a LOT more of the original, but images have a much much wider variance unless you ask specifically for as many keywords as possible (aka Mona Lisa by Leonardo Da Vinci) in which case you'd fall under trademark law as well probably (Which is also why Microsoft added trademarked images on their CDs so they could prosecute infringers easier).