Picking the best best paintings requires artistic judgement.
So I mistook your analogy. You cannot equate the output of gen ai with the Louvre. The paintings in the Louvre are there for historical reasons and because people that claim to know art, have already selected them for being famous. They are not there because of randomness, so that any thief would have to use artistical judgment to pick the ones that look better for an intended purpose.
Surely you can see how the way the photo looks depends on how the statue looks? The sculptor spent months working on the sculpture. The photographer spent maybe an hour.
So? Both of that is not relevant. It is different art. You cannot compare them directly. You do not chose between the sculpture and the photo. You chose between different sculptures and between different photos.
You sound like someone that thinks cooking food is just putting everything into a pot and, well, cooking it. After all, the ingredients are already there and have nutritional value, no matter who cooks them. So we should not praise the cook at all for the skill involved. Instead we should only praise the farmer for the taste.
A good photographer will make even a bad looking statue look good on photo. That's what the skill is. And not the skill of forming a statue. That sculptor will probably make a bad photo of the statue. And the photographer probably cannot sculpt. And taking photos of statues, without further context, sounds more like work and not like art. Hunting statues in the wild that are withered and taking photos of them in the right conditions to create a picture book on the other hand ...
Anway, photos of sculptures are established art activities. Go to them and explain them, that it is not art what they do. I do not need to convince anyone here, not even you. The real wold of art already disagrees with your opinion about photographies of scluptures and their artistical value.
But does he deserve the top billing he usually gets? Not really.
So you are consinstent. I appreciate that.
But you do not deny that the director has had artistic input into the work, do you. Because now you are just haggling how much input it was. And that's a moot discussion. You cannot compare the artistical value of a director to that of a screenwriter or an actor. It's too different. And for a work of many different artists, you can not really quantify it. It's not unsimilar to cooking. Just putting known actors and screenwriters together will not make a good movie. You need a force that orders all those things together. That's usually the director, but movie making is complex, so it might be other people too.
For gen ai, the use of a prompt often does not fulfill the modicum of creativity, the threshold of originality. Or in the analogy, it would be someone just pressing their phone's photo button to take a snapshot. I think, it is possible to create art with it. But most things one does encounter will not be such. Just like most photos are just selfies and snapshots taken by amateurs. And like a lot of actual art is just pretentious garbage, in the eyes of many.
And I do think we need to let go of the plagiarism argument. It would fall flat, if you had a llm model with explicitly consented training material. But you could still do all the things you can do now with gen ai. Would current ai opponents suddenly cherish ai works, if the model would be ethical? I bet not. Focus on copyright and "theft" is short sighted. And using terms like stealing, while it is just unconsented training of a machine at worst, makes the arguments sound like hate, instead of arguments. You will not convince people this way. Just like you will not convince meat eaters by calling it murder what they do.
I think a better approach is to use preference for human works. Because that will stick, even when there would be low energy ethical llm to produce gen ai.