The argument that it's just mashing up existing art is odd anyway, since it doesn't (normally) actually just take specific pieces of other images and form some kind of Frankenstein collage. It analyzes patterns and then tries to auto-generate something that fits within those patterns.
It also often does a terrible job.
I spent way more time than I wanted (and eventually gave up) trying to get DALL-E to generate a reasonably good dragon yesterday. It gave me a lot of versions of what I was asking for but they all sucked.
It would have been a lot easier and better if it actually just searched Google for "dragon" images, picked the top results, isolated the dragon from the bg and gave me that. But that's not how it works.