House Styles
Text-generating AIs get a lot of attention for their (apparent) capacity to outperform a human subject at any number of useful analytical tasks. An equally demanding challenge might be to persuade an audience that AI-generated ‘artwork’ was produced by a human. All the image generators I have seen might be skilled in some realistic or technical sense. Still, by now the visual tip-offs that an AI was involved are increasingly well-known. The specifics differ somewhat by the model, but the overall AI ‘house style’ for visuals is–to my eye–quite recognizable and consistent at this moment. Harder to explain is this: how does one recognize a piece of art as the product of a human? The human is what is left when all known AI styles are subtracted from the world.