Could AI Appreciate Art?

A text and exhibition exploring the possibility of AI art curation

Adapted from “A Human Exercise In Computer Vision” that served as the introduction to a curatorial project by Loraine Wible, Lindsey Whittle, and myself… in collaboration with AI.

A work of art by Lindsey Whittle (left) and a AI-generated text (right). IMPORTANT NOTE: the AI-generated text was lightly edited by a human for clarity and conciseness. It was generated by a GPT2 model from a human-written prompt summarizing several Computer Vision model assessments of the image. This text was chosen by a human from around a dozen responses.

Could an AI understand or appreciate art? Is it possible for a computer to form a critical opinion? Or have an aesthetic response? What does it “see” when it looks at a work of art? To explore these and other questions, we enlisted an AI (or, more precisely, several AI models) to co-produce this curatorial project.

Front page of the PDF version of A(artistic)x I(nteresting)x P(owerful).

Which is… what, exactly? A book? An exhibition? It doesn’t fit neatly into either category or any of the common intermediaries (artist book, exhibition catalog, etc.). Complicating matters, there are no images anywhere, either in the gallery or the PDF. It’s certainly a text, an embodiment of an experiment, and a proof of concept (more on that later); a version of this PDF’s contents were installed on the walls of the Pearlman Gallery at the Art Academy of Cincinnati in late 2020, so it’s also an exhibition. And if you have the means to print out this 340 page document and enough wall…