1992, Poet Anne Carson has published a little book called short talk. This is a series of miniature essays, from one sentence to one paragraph in length, on seemingly unrelated topics – orchids, rain, the mythical Andean llama. Her “brief talk about what it’s like to take off a plane” is what it sounds like. Her “trout short talk” is mainly about the types of trout that appear in haiku. In the book’s introduction, Carson wrote with dry Canadian relevance, “I’ll do anything to avoid boredom. It’s a lifelong mission.” It was around the time she published the article that the internet took off.
Fast forward 30 years, and at least for me, one of the latest ways to avoid boredom is to stay up late and play with AI image generation. Tools such as DALL-E 2, Midjourney, and Stable Diffusion can be instructed by text prompts to create a Titian-style faux oil painting of a dog in a hat, or a simulated photo of a plasticine model of an astronaut riding a horse. When I first started using Stable Diffusion (open source and very interesting), I was reminded of Carson’s talk. I went back to them to find out why.Soon I realized the similarities with form.
Everyone says content is king, but the secret monarch of the content economy is form—constraints, rules, minimums, and maximums. You grow up learning form. High school composition is five paragraphs. The sitcom has eight minutes of commercials in half an hour. Novels are long. Tweets are limited to 280 characters.
How is my tweet, essay or studio film different from yours? The choices each of us make in the table.In a word, our style. Carson’s book takes a familiar format, small lectures, subverts it, manipulates it until you as a reader start to feel like you’re in her wonderful brain, scrolling through her mental browser history, joining her hyperlinked Fantasy and Semi-Abandoned Rabbit Hole. Image generation is a bit like — but instead of talking to a smart Canadian brain, it’s talking to a giant idiot world brain. (A less neurotic way of putting it: A multitude of data objects grouped in layers, connected together in incomprehensible degrees, like art on a multi-masted sailing ship’s rope and nails wall, but set on fire with the flow of data.)
In general, humans like to use machine learning to help pathologists, sharpen cell phone photos, or make better maps. But AI generators will get a lot of attention. These tools work by scraping images from the internet, absorbing the visual culture contained within them by scanning their captions, then adding frothy visual noise to them until they appear static. To make a new image, the AI starts with a title and some static, then runs the process backwards, removing noise, until an image emerges that more or less aligns with the title. (Not a good painter, but so am I.)
This feels disgusting. It’s disgusting to see the artist forgotten. Someone can say to a computer “I want a Frida Kahlo-style portrait of Alex Jones” and it feels gross, and the computer will do it without moral judgment. These systems roll scenes, territories, cultures—what people consider “theirs,” “their lives,” and “their crafts”—into a 4 GB open-source zip file that you can download to Mac in order to make a Miyazaki-style baseball penguin. Those who can use new tools will have new powers. Those who are good at old tools (brushes, cameras, Adobe Illustrator) will be thanked for their service and rendered into Soylent. It’s as if a guy in Allbirds stumbles across a neighborhood, and everyone grapples with it and says, “I love this place, it’s so quirky! Siri, play my Quirky playlist. And open one around the corner. Blue bottle!”
So naturally, people terribly upset. Art sites are banning AI-generated works, at least for now. The stock image service also rejects it. Prominent bloggers who tried to use AI to illustrate their writing were punished on Twitter and pledged not to do so again. AI companies talk about ethics, which always leads me to suspect that certain words are forbidden in the interface of the image generator, which is sad because I want to ask the bot to draw a “plump” hut in the style of Thomas Kinkade. (One must confront one’s own deepest fears.)