

not because they’re ever good but because of the conditions they were made under. So, still actual art.
I’d argue that torturing GPUs into generating thousands of extremely similar but still unique pictures of anime titties has a bit of poetic artistry to it
(I’m mostly shitposting here, but there might be a kernel of truth somewhere in there)
Except you can very much use it as a tool to create a work. With img2img you can feed it a drawing (or anything really) and have it remake the image you fed it in a whole new style. With inpainting you can take an existing image and have AI fill in a masked area with something else entirely. Controlnet lets you control how things in the generated image are composed or even do wild things like create unique QR codes. And absolutely nothing is stopping one from taking the AI generated image, pulling it into an editing software and editing it further into a greater piece too
With tools like these its very easy to see how an artist might make something like say, a fake time travel log or use it as a tool to reduce how much labor is needed to complete a larger work
AI is like photography. If all you do is rely on the machine you get output exactly as good as the thought and effort you put in, but if you approach it as a tool to use in your artwork it opens up a ton of neat possibilities that are yet to be seen and can be a brilliant tool to keep in the toolbox