Pathetic Fallacy

If one can set aside the many, many ethical issues involving AI art, the question remains for its detractors (like me): what exactly do its proponents see in it? It’s certainly gotten “better,” in the sense that the people in the images now usually have the right number of fingers, and any text isn’t a garbled mess of ersatz letters, and there are fewer instances of what H.P. Lovecraft would call non-Euclidean architecture. But for me at least, the more accurate AI gets, the less it appeals. In the early days of DALL-E and Midjourney, when we were all stuck at home in the midst of a pandemic, typing silly prompts and getting smudgy blurs in response was a lark. Like the joke about the dog who could only type 20 words a minute, the remarkable thing was that it worked at all, not that it was any good. But today, when I prompt Chat GPT with “patriotic American family with a boy and a girl and a dog watching TV,” the elaborate tableau generated is so on the nose, and yet so soulless and dead-eyed (see below), that it renders all the state-of-the-art computation (and the presumably extravagant energy use) irrelevant. All the figures face forward and arranged using strict isocephaly; the virtual canvas is arranged with horror vacui that makes Where’s Waldo look open and airy.

There have been a lot of think pieces about the politics of AI art, from claims that it’s beloved by fascists, to the counter belief that it’s a great democratizer. It does seem to have a particular home on social media, Facebook especially, where AI’s prosaic manner lends itself to oversimplification, cliché, and moralism. When you don’t have to do the work of actually visualizing what you’re saying, you don’t have to make sure your ideas make sense and that your facts are, well, facts. Part of the process of creation is realizing that your visual problems may actually be conceptual problems. Similarly, viewers looking for confirmation of their own beliefs are more easily swayed by images that have the veneer of reality. Or they might just accept them as real.

Whatever the politics, it’s this effortless, cut-and-dried nature of AI that makes it so tedious. Allow me to make another one of my patented far-fetched analogies. In the mid 90’s, during the heyday of Microsoft Office, clip art was everywhere, and the most overused clip arts of all were the Screen Beans, a series of bulbous human-shaped silhouettes doing things, or more commonly, reacting to things. These illustrations were designed by Cathy Belleville and licensed to Microsoft for distribution with Office in 1995. They depicted poses that were purposefully vague, so that they could be used in any situation; however, that vagueness also drained them of any meaningful content or personality.

But, boy howdy, they got used. In Powerpoint presentations, yes, but also in church bake sale signs and guitar lesson flyers and passive aggressive notes reminding people to pay into the coffee fund. This was before most people had Internet access so it all got printed on the sly using the office laser printer. So many trees gave their lives for the millions of reams of 20 lb. copy paper that were emblazoned with a screen bean jumping in the air or scratching its head in bewilderment. But for all the ubiquity of these inky nebbishes, they never really gave the texts they accompanied any new information.

So why do these generic, overused illustrations remind me of AI art, which is supposed to be bespoke to the user prompt? Both are art for people who really don’t care about art. They are perfunctory nods in the direction of art employed by people who lack the skills, funds, or interest to do better. Clip art, like AI, was presented as a democratic form bringing design to the masses. Why limit art production to people who spent their lives developing a skill, who expect to be paid for what they do? This will do instead. But to paraphrase Johnson, “what is drawn without effort is in general viewed without pleasure.”

The history of art has been the history of its production and distribution. When books had to be written by the few who were literate and copied painstakingly by hand, there were few books, but they were highly valued by writer and reader alike. Similarly, music production once required musicians who had invested years into their craft, as well as had access to instruments or could make their own. Listening required finding these musicians, organizing them, and gathering an audience. Painters had to apprentice with masters in their workshops; they had to know how to mix linseed oil or tempera with rare pigments. As people learned ways to mass produce their tools, to replicate their creations, and to widely disseminate the results, the arts changed. And this was a good thing, because it meant greater access for art lovers, a lower bar to entry for potential artists, and less cost for everyone. Technology in this case really was democratizing. But up until now, however it was made, the creation of art had to be intentional, and took time and practice.

This is what’s lost when effort is eliminated. Whether you mine your own cobalt to mix your own paint or you draw in digital media on a tablet, the effort is the art: not just the act of creating, but your motives, your lived experiences, and your personal aesthetics are the ultimate media of your work. Likewise, the effort an audience brings to close attention, to interpretation, to contextualization—that’s the other half of art. And if we give the production over to machines, we may as well design an AI to enjoy it.