AI art is irrelevant, and I can show you why using a simple analogy.

Digital art is essentially a grid of pixels, and each pixel has an individual value, i.e. a tuple that defines the various color and alpha channels. Another grid of data is an spreadsheet. In theory, you could train an AI on a bunch of spreadsheets with the same structure, and then ask it to create a spreadsheet based on that training.

Here’s the issue: what would the point of the output spreadsheet be? When you create a spreadsheet, you’re taking data that is relevant to the situation and placing it in the grid in a structure that makes it simpler to understand a problem. The data in a spreadsheet is only relevant in the context of the real world situations represented by the data. Without the binding of data to human experience, it’s just numbers that are highly irrelevant. A predicted (generated) spreadsheet is worthless.

Similarly, a piece of visual art predicted by AI is also mostly worthless. Sure, it can depict something that you might expect from the data set of images the AI is trained on. That image, though, is going to be disconnected from the lived experience of both the artist and the audience. At best, you could hope that some kind of abstraction of that experience is somehow transmitted through the training, but it will always be useless data. The predicated image can’t be relevant to lived experience, just like the predicted spreadsheet won’t have anything useful to say about a problem you’re trying to understand.