I look at generative AI the same as any other kind of tool that an artist can use as part of their workflow. Computers have been automating large parts of the creative process for years, with tools like Photoshop filters, Blender scripts, Substance Designer nodes, etc. Naturally, some programs like these are starting to implement AI-driven plugins. I think using AI tools is perfectly fine as long as they're used as just that - tools. I don't see much value in trying to sell or expect credit for something where an AI has done the majority of the work. That would be like generating some Perlin noise in Photoshop and putting it up on an asset store.
As far as the issue of plagiarism goes, I think it really depends on the individual images. The way people discuss AI, it seems like there's this assumption that if copyrighted artwork is involved in the generation process in any way, then the output necessarily must be considered plagiarism. I don't believe this use of the term is appropriate. The first requirement to claim plagiarism is that the art actually looks like an existing piece of art in the first place. Human artists are influenced by other art all the time (it's impossible not to be) but that alone doesn't make the results of that influence plagiarism. You have to demonstrate that what they made bears meaningful resemblance to another author's work. By "meaningful resemblance" I mean elements that are unique to the original author, and not just "both images are a picture of an apple".
Is AI guilty of this sometimes? Absolutely. As mentioned here before, some models are capable of reproducing artists' signatures, among other things. In these cases I believe it's the responsibility of the user not to use such content in ways that would infringe upon the rights of the persons being imitated.