I don’t agree with this argument at all, because if a human artist were to employ the same kind of algorithmic mimicry that an AI does, I would consider it plagiarism. There is a distinct difference between how a human observes and learns from other artists work, and how an AI does it.

Moreover, to take things out of the realm of plagiarism, if a human artist was mimicking the style of another artist and making bank off of it, and the original artist were to say “hey, that’s kinda not cool, I don’t appreciate this” you could have a conversation about how to accommodate both parties. With AI, there is no such conversation to be had, because it will replicate without barriers and do so in volumes that dwarf any sort of output the original artist could dream of, no matter how nicely you ask it not to, unless it was not trained on it in the first place.

Anyway, my pushback in my original message was not about the output being plagiarism or anything of the sort, it was about the usage of authors/artists work as training data (input) being non-consensual.