Digital Art and Human Development

Art produced by Dall-E (my collection)

Friends widely embraced the Lensa AI art app this week for one simple reason: it could create faces. This year has seen remarkable advances in computer-drawn art, beginning with Dall-E, which took social media by storm this summer. But Dall-E couldn’t draw faces. Human likenesses came out pinched, distorted, with nightmarish proportions. Dall-E’s immediate successor, Midjourney, did better faces, but they were cartoon-like and whimsical.

Lensa, by contrast, could read users’ selfies, and the program could return artistic renderings. Though the pictures aren’t yet realistic, the computer was capable of making value choices which presented finished portraits that emphasized certain features like trained artists would. My friends uploaded their selfies, knowing they were relinquishing rights to a massive AI clip-art pool, because the reward was painterly renderings of themselves as wizards, spacemen, and cats.

Pushback began almost immediately. Not only did Lensa pinch your uploaded images, the claims read, but it also pinched legitimate artists’ hand-made work. The app created painterly renderings because it modeled itself on actual painters, the techniques they used, the decisions they made. AI programs are already putting human artists out of work, though not yet in huge numbers. The ethics were already creating headaches, and it’s likely to get worse.

I’d like to focus on a different ethical consideration. Yes, the human economic challenge is serious: just as American industrial jobs moved to Japan and China, where labor was cheaper and supply lines were shorter, AI art is likely to displace human artists because the renderings are nearly instantaneous. People are unlikely to pay more for slower human-made art just because doing so is right. But the problem begins much earlier than the point of sale.

Art produced by Midjourney
(Wikimedia Commons)

Many schools begin teaching art in preschool. We teach children to draw with charcoal and Conte crayon, to play “Alouette” on a plastic recorder, and to sing in unison, because these represent important developmental milestones. Art and music teach children important skills of hand-eye coordination, the ability to sit still, and patient dedication to tasks where the reward isn’t always immediately obvious. Children need all these skills in multiple disciplines.

As I’ve gotten older, though, and spent time on both sides of the classroom, I’ve realized students learn even more from art. When drawing, students learn what to include, and what to omit. The entire process begins by reducing the forms we’re rendering—landscapes, buildings, human bodies—to a few essential lines. Then the artist makes choices about which shapes, colors, and textures to incorporate. These are important value choices.

Anybody who’s witnessed the products of a freshman life-drawing class knows that a roomful of art students will produce very different images of one model. Photographers may get slightly different images from the same model, depending on choices like shutter speed, angle, light saturation, and Photoshop aftereffects. But human artists, looking at the same model, can produce wildly divergent paintings of the same base human form.

These choices are subjective, but not value-neutral. The proportions of the human form, like the proportions of musical notes in a composition, have mathematical relationships, and artists make choices about how much they value harmony, synchrony, and continuity. These choices speak to the artist’s inner process. But they’re also external, and speak to what kind of reaction the artist hopes to produce in the chosen audience.

Art produced by Lensa (promo image)

Someone must make similar choices with AI art. The person writing Lensa’s prompts, of course, chooses whether they like the image produced. But some programmer deep within the bowels of a Silicon Valley industrial park also chooses what proportions, textures, and colors the program will emphasize. That programmer made these choices months, even years, before the artwork was produced. But that person made these choices, on your behalf.

We already know that recommendation algorithms on streaming services like Netflix and Spotify both empower and circumscribe our choices. We receive access to more musicians and filmmakers whose work we might never have discovered. But we also wind up encountering fewer kinds of artists, and wind up watching or listening to the same genres repeatedly. Our tastes become deeper, but narrower.

If programs replace artists, which seems likely, these AI art generators are likely to have similar effects on students’ ability to make choices. As art becomes instantaneous, and not a process, students are likely to truncate their ability to even see the choices being made, much less implement them in their own work. Their worlds will become narrower. If choices are instantaneous, why make any choices at all?