It doesn't change the fact that "gets trained off of the work of human artists" is how actual human artists learn their craft. 

As to any actual good, while I've seen a lot of PR about the potential of AI in medicine, science, etc. - ie, the things where there would be real, tangible benefits, notably none of that has come to pass. I suspect this is because AI (in its current form) doesn't really ideate. It's maxing out one core aspect of human evolutionary success: pattern recognition. Great for sorting data, terrible for analyzing it.

I'll be perfectly happy to switch my view on it, since the creep of AI into all corners of human endeavor seems inevitable and I'd rather not be walking around feeling glum about it all the time (there's really so many other things to feel glum about). But I need proof that it is a beneficial tool, and to get that proof it needs a big, irrefutable, and applicable win.