When I was young I thought photographers just took like one photo and by sheer willpower and skill, they sometimes captured beautiful scenes. I then found out that they actually sit there and click the shutter thousands of times, then they go somewhere else to sit and scour through the "generated art" to find the gems. Then they polish the gems. Then they publish those.

My novice AI-art workflow involved a similar pattern. I come up with some topic or thing I want to see, then I iterate on the wording of the prompt and all the little variables the tool takes. I end up with dozens of crap pictures, but sometimes cool stuff comes out. If I keep working on this, alongside increased generation speed, I could see myself "creating" really good art.

To me it is about the process. It's the human input that makes art valuable. I mean, if AI becomes sentient then those judgements would be a valuable input in themselves too.