The case is slightly stronger if you want to talk about training that happened on our private code repositories, but publicly visible art and repositories are fair game for learners to learn.
(to be clear, I don't think this is an accurate description of how artists feel about e.g. publicly released diffusion models. The core is not that it's made by "corporations" or the for-profit aspect or even the copyright infringement part, it's simply the understandable and not-at-all-unique-to-artists resentment at having your job or passion automated or displaced)
Why would anyone be fine with people explicitly not actually making art.
Scrapping and generating isn't art and isn't comparable to people tracing/copying/learning from existing art.
We've been automating people "out of jobs" since at least the start of the industrial revolution. It's generally considered to be a goos thing.
We had a sort of weird special case for copyright preventing certain types of machine automation and I've heard plenty of complaints about how that worked, but it more or less did work for a while.
By who? The corporate bag holders?
I don't know the details of ceramic industrialization.
I could sort of get behind something like this in theory (weighing each new technology based on whether it furthers the values of the community, and rejecting it if it doesn't, kinda like what the Amish do), but I'm not sure it's being applied neutrally or consistently here, nor do I think it works in practice especially for something as decentralized as a diffusion model, which are surprisingly simple to build and train and can run on a home GPU.
...yet. AI has evolved greatly from basic image generation. It's already pretty capable at using tools to great SVGs and 3D scenes that are more reliable than plain images. Sure, experts won't lose their jobs, but how will juniors become new experts?