I don't think AI art is art at all. And before you start labeling me an art snob and dismissing art as "irrelevant", let me rephrase that - I don't think that AI images are as important as human-made images are. I have a Master’s Degree in Art History, so bear with me here.

For humans, images reflect what we experience with our eyes, which are part of our biology. Making images such as paintings or drawing reflects our gestures (brushstrokes, pencil markings). Most human art is “human scale”. Think of the size of the canvas. Even video and film reflects the eye of the user and their vision, although now they have just been reduced to an eye (I think Jean Paul Sartre wrote about this, or maybe Jacques Lacan in "The Eye and the Gaze"). And our paintings, which often represent bodies or landscapes (decreasingly so after the 20th century) tie in materially to our own human embodied experience.

In “The Beauty of Everyday Things” Soetsu Yanagi wrote about the superior beauty of craft, which bore the imprint of the human hand. And that is why the best crafts, in Yanagi's view, have some imperfection. I’m talking about wabi-sabi here, the chipped bowl that bears the marks of many years of use and becomes beautiful, or the Velveteen Rabbit. Soetsu Yanagi gives the example of the process of creating Kasuri indigo cloth, which had been done for many years and was shaped by generations of human hands until it became second nature, almost a natural process.

At one point, there was an attempt to perfect the cloth dying process so it would be flawless for the Japanese royal family. The result was technically flawless but also depleted of its natural beauty.

Like the Kasuri cloth created for the Japanese royal family, AI art bears no evidence of the human hand. It is a performative imitation of human art. AI images often represent something that resembles the human form, but what is it? It’s an averaging of lots of data about what human bodies should look like. AI “art” is more like a chart than an image. It's like the image of the beautiful Japanese girl that was actually a hologram, in William Gibson's book Idoru. It is definitely not a vision generated by the human eye.

But who cares right? People are having fun generating pictures of themselves as AI sees them, and it's kind of funny in an uncanny way. It's harmless, right? Maybe. Recently I saw a talk for a local historical society. The topic was related to the Semiquincentennial, or USA 250 years of Independence celebration. The speaker relied almost entirely on AI-generated images throughout his presentation. Did he realize that there are many, many examples of 18th-century American painting showing the exact events he was describing, executed by humans? Maybe he did realize and instead wanted a collection of images that looked like posters for the Terminator.

If you want to find something like AI art, don’t look to images. Look at its code output. Code is to AI what images are to humans, an outward expression of their internal composition and experience. AI code is imperfect. It’s often tangled and overcomplicated, but also the most authentic version of what AI is, its essence, and how it sees the world.

So I think that code is to AI is what images are to humans - an outward expression of the internal composition and state of its maker.