The point is, there’s no print that is 1:1 equivalent to an original painting. There are artistic choices embedded throughout the process that influence the result and make one print different than another, which people might consider more or less accurate, or faithful, or beautiful, or artistic.
It’s easy to assume an image is just an image, a bunch of pixels with defined numerical values. But the real world has so much more detail than that.
Basically, art speaks to our sensibilities, so we expect to find a message in there. LLMs can't create, only mix-and-match what exists in their training data set. You could slop a new painting, make a special canvas printer, and exhibit it in a blind test. Some people will think it's art because they find meaning in it, while others will think it's not art because they'll recognize several meanings slopped together in a nonsensical way.
In reality, it was never art, because the creator, a statistical algorithm, never had any message to convey in the first place.