Will AI-artists make humanity less creative?

I find this take elitist for one, and just plain short-sighted overall. It does raise some good questions though.

First of all, it's clearly a net better for more people to have access to the ability to create beautful things without having to invest years of practice, the speed of which is limited by our biology. So more people can make cooler stuff with less effort. As for whether this cheapens the process, I suppose it does by definition but I see it as a good thing.

But I guess the real question being asked here is if it reduces the incentive to develop skill. And I'd say yes and no, it probably reduces the incentive to learn to draw by hand, for example. But it also now forces artists who're enthusiastic about art to take full advantage of these tools. What I expect we'll end up seeing is that the standard just gets raised massively because these top 1% of artists will now be the top 1% of people who can make the best art while taking advantage of these tools better than everybody else.

As for the argument that it's no longer the artist who creates the art but the program, this is a misunderstanding of the creative process, which consists of an extremely long chain of creating and criticizing. DALL-E just makes the creation bit a whole lot easier, but at the end of the day, it's the artist who decides if it's good enough to consider finished, and ultimately people who judge it based on their own artistic preferences.

Just to give an analogy, is it sad that most of us take notes and write entirely digitally? Have we lost or at least marginalised the creative practice of typesetting and calligraphy because of latex? Yes, but that means more people can now afford to focus on other more valuable things, like writing research papers. All the while though, value is still judged in the frame of reference of other people.

All that said, I think there is one real concern here. That these systems get such a deep understanding of human preferences that even the best artists can find no flaws in the first iteration of whatever it is their AI system generates. But I don't think this is very likely. This would mean each AI would need to have such a deep understanding of each artists' preferences that it can game them easily. I believe this would need a complete model of that human and all their experience, and at this point it'd be at the level of a digital clone, which I guess doesn't sit well with some people, but this level is probably a ways away.