Raw LLM Responses

"It doesn't work" "Prove it" "That's not the point. If I don't say it doesn't work, you'll keep trying. Even if it doesn't work now, it will work eventually, and I'll be screwed, because I already poured a ton of money on those companies." Also, the argument of "it can detect poison" is bullshit. "Detection" isn't simple, and these models don't have certainty. It can guess, and have a rough idea of how good that guess is, but still, best case scenario, it will "unpoison" regular images too or stop using valid images, which is a blow to the models anyways. Keep the poison going. Another note, control and automation engineer here, who understands neural networks and how they work (calling it AI is just marketing to make it seem smarter than it is). Calling neural network models data training "inspiration" feels wrong. Inspiration is about distilling the information, breaking it up, deconstructing it, and rebuilding it along with your capability and experience. You can only reproduce 1:1 if you want, but artists will never do that (unless for personal training). These models straight up add up the images to the models, just instead as data. While it doesn't straight up store the full unmodified data, it stores as close as that, as you saw with those examples. It's not inspiration, it's copy. The only thing impeding these models from taking copyright strikes is the ignorance of lawmakers, or money from these companies paying politicians via lobbying. After these companies lose enough money and/or enough lawsuits are processed by juries, situation will reverse. Like the Sarah Andersen, Kelly McKernan, and Karla Ortiz lawsuit, that already started on a bad foot with only a few acceptances on the claims and awaiting another trial after the claims are updated and discovery is completed. If that judge isn't bought, maybe that one will do something (after seeing these outputs that feel more copies than "inspirations", you'd have to be blind to not see it as copyright infringement), but I trust the US legal system as much as I trust an untrained famine dog near food.