#would an algorithm be able to reliably predict which paintings fell into which category? would a person? does ai art consistently pass the

[thefictionshelf]

listen I’m not really qualified on A.I or art so maybe this is a dumb take but like. I think the fact that the AI essentially takes the average of a mildly curated huge database (and the fact that those same A.I generated images will end up in those databases) of images is going to make the technology a lot less job-destroy-y than (some) people are predicting. I’m not saying it won’t shake up the industry but I think that A.I art, especially for vague inputs, and stuff that people are making a lot of, is going to start to have a look. The first time I saw one of those AI fantastical cityscapes/libraries (you know the ones) my socks were knocked off, the second one was pretty cool too... now? I mean they’re pretty... but eh. This might totally be my internal biases, but i’ve found repetitive A.I art to get boring in a way that I haven’t from human artists, even when they’re repetitive in style and subjects. There’s still a freshness to it.

There’s a photographer I follow on instagram, and they’ve started this series where they take a photo, ask an A.I to expand it x4 and ask the audience to guess which quadrant is the original picture. and you know what? it’s really easy. Not because the generated portions look unrealistic, or uncanny, or because there’s some awkwardness at the seam between the original picture and the newly generated one. That’s all flawless (at least to my untrained eye.) It’s that the human taken picture is always the most interesting part. The most unusual, strange, unique, new, challenging. It always ends up drawing the eye, becoming the focal point, even when it’s not as busy as the rest of the photo. Three quadrants of these photos are clearly mined from the average of a million ~aesthetic~ pinterest photos, and one is the result of a strong, bold, consistent creative vision. 

If you’re a company and you want a quality, (sometimes they care about that) brand recognizable (they always care about that) piece of art, it’s probably better to start with a human artist. I’m not saying that there’s no room for A.I and for those who know how to use it in the process, but I can’t imagine constructing a truly strong, unique, piece, from a game asset, to a logo, without human input at multiple points during the process.