Very useful: in the future, we want to have machines that can make the same (peer-pressure-free) art classifications that humans would, so they can pop out art themselves. Bad art is nearly as useful as good art in helping to train such a machine and identify the algorithms humans use to make these judgments.

But when the field of art has been corrupted to the point where it's just a pure status game, there is no such classifier that can be learned. The only machine you're going to be making is one that looks human, and hobnobs its way up the social ladder so that it can learn what the elites think, and render judgments in that way.

(I made the same critique about some Japanese researchers' quixotic attempt to build a machine that determines how much humans will like a given wine, based on chemical analysis. Hey guys -- it ain't the chemical composition of a wine that makes people like it!)