AI image generators like Midjourney or DALL-E are generally adept at capturing the accuracy of the human form. The concerns over copyright, job infringement, and general degradation of the visual arts via such AI are ongoing concerns for many artists and practitioners. However, a new New Yorker article by Kyle Chayka identifies a noticeable flaw in AI artwork: human hands.
Missing the Big Picture
Chayka recalls an art class where he was asked to draw his own hand. It’s an assignment for beginners, and as behooves a novice, tempts the artist to focus more on the specific contours of the hand instead of the overall structure and form. The forest gets lost in the trees, so to speak. AI is guilty of a similar flaw. In many artificially contrived images, the hands come up gnarled, disfigured, or otherwise anatomically incorrect. Some of them are minor yet noticeable mistakes upon investigation. Others are repulsive, looking alien and mutant. Chayka writes,
A generator can compute that hands have fingers, but it’s harder to train it to know that there should be only five, or that the digits have more or less set lengths in relation to one another. After all, hands look very different from different angles. Looking down at my own pair as I type this on my laptop keyboard, my fingers are foreshortened and half obscured by my palms; an observer wouldn’t be able to determine their exact X-ray structure from a static image.KYLE CHAYKA, THE UNCANNY FAILURE OF A.I.-GENERATED HANDS | THE NEW YORKER
Hands are hard for computers to replicate given the many angles they can rest in and their general complexity. The result isn’t pretty.
Chayka thinks that someday we will look back on this flaw with wistful nostalgia. Eventually, AI will learn the form of the human hand and no longer be a novice in Art 101. We will then wish for the days when it produced a bad hand and could remember how a real person could do it more justice. For now, though, AI’s failure in this arena shows a gap in its capacities and highlights an area in the arts still best left to human creators.
Art and Human Bonding
This week, professor of economics Gary Smith wrote on the importance of critical thinking and writing skills for students. ChatGPT, he says, tempts students to outsource their cognitive brainpower to the machine, but notes that this is setting kids up for failure. In addition, ChatGPT will fail to perform complex written tasks that are best left to humans. He also notes that “writing bonds us,” commenting,
When I tell you what I think, you learn more about me. When you respond, I learn more about you. We learn about our similarities and differences and, if done politely, become closer. All of that is lost if our written communication becomes my LLM chatting with your LLM.GARY SMITH, LEARNING TO COMMUNICATE | MIND MATTERS
I wonder if Smith’s wonderful observations here might translate into the issue of AI “art.” Both writing and drawing depend acutely on observation. We have to pay attention to the world around us before we can link together a coherent sentence or draw a glass vase sitting on the kitchen counter. In addition, what if art, like writing, “bonds” us? Isn’t that, after all, one of art’s principal purposes? Even if Midjourney and DALL-E master the human hand and its images become indistinguishable from the best digital art done by humans, it lacks a human genesis. It fails to “speak” to us.
Communicating with a Live Intelligence
The essayist Becca Rothfeld frames the problem in terms of whether ChatGPT could write a novel, writing brilliantly in this piece from The Point,
Indeed, we read novels, rather than textbooks or user’s manuals, because we are not in the business of extracting propositions but in the business of effecting intimacy with another live intelligence. Literature is not (only) a conveyer of information but a locus of communication, and we cannot communicate with an inanimate mechanism, whirring its insensate way through text it does not even comprehend. For this reason, we could only ever really care about words that have been deliberately placed on a page by another person. Books, the German Romantic novelist Jean Paul once wrote, are “thick letters to friends.” Who would want to correspond with the void?BECCA ROTHFELD, WHAT CHATBOTS CAN’T DO | THE POINT MAGAZINE
In Rothfeld’s estimation, the technical abilities of AI don’t lay a foundation for its artistic and linguistic value. We don’t go to books and art to ingest information formulated by “the void,” but, as she so eloquently puts it, to commune “with another live intelligence.” The more we value literature and art created by humans, for humans, the more we can discern AI’s proper use and purpose.
Cross-posted at Mind Matters News.