Raw LLM Responses

As someone who does research in Machine Learning and Attacking Models and planning on involving it in grad school I wholeheartedly agree with your opinion. I do differ on some points but all in all. I think that removing the human aspect completely makes it devoid of what art can be. As a musician and an artist, The experiences and meanings behind a piece are just as important and you are well within your right to protect your craft. I have seen some good uses for AI art, but never as a final product, I myself liked the older models that broke and just went crazy because through that random mess they produced you had the chance to capture interesting things. Like models don't have rules in their mind, so you can definitely use it as brainstorming. For example some bugs have been used for making interesting models there is one that sort of makes everything look like hallucinogenic with many eyes which is very uncanny. But outside that I wouldn't use it for a final product at all, more less so with what people are doing nowadays which is just trying to replace artists. Edit: In regards to the poisoning of data yeah it can be defended against, but in my experience this being a relatively young field, you start to get diminishing returns, you start to sacrifice some performance on the normal clean data to resist the poisoned data. And even then most models might not make defenses for it. That takes time and effort and well it might just not be a worthy venture to a certain degree. Now for model collapse, while yes it means that newer models will suffer and not generalize better compared to their earlier counterparts the problem remains (there are models with good data out there) so AI art will still be a problem even if their newer counterparts suffer from model collapse. And that does not remove the ability to try to find new architectures to create said art. But its good to poison the data, and create a movement. Artists need to be protected.