AI and Art and Pettiness
Two new events have triggered another round of gleeful insistence that imitative AI really is creative and just as good as humans. I am unconvinced, and more, I am quickly losing respect for the people who make these assertions.
I really don’t want to get too deep into this conversation, because I think that others have already said what needs to be said. Noah Berlatsky points out that most people (raises hand) don’t like poetry and so:
The human truth that the AI is reflecting is that people don’t like poetry. Of the 800 million books sold in the US, only about 3 million are poetry. A successful fiction book can sell millions of copies; a wildly successful poetry book will sell a few thousand copies at most. The vast majority of people in the US will never buy or read an entire book of poetry in their lifetime. Even regular readers almost never read poetry; if you have read a book of poetry in the past year, you are a poetry expert with vastly more knowledge of poetry than virtually anyone you interact with online or in person.
…
Poetry is so utterly alienating and uninteresting to most people that virtually anything that looks like poetry but is not actually poetry poetry is going to be massively more popular than poetry. You can go down a list; rhyming children’s books sell more copies than poetry. So do hallmark cards. People who would never buy a poetry book or read even a short poem will recite entire albums of pop music and hip hop lyrics from memory.
Max Read points out that the art study was almost designed to fool readers:
One obvious problem with the experiment is that Alexander has stacked the deck. The test is effectively designed to fool people, as Alexander admits--the “human” and “A.I.” works are in each case being chosen as to not demonstrate any of the features distinctive of human or A.I. authorship, among them “text… complicated wrestling-like poses… and pop art,” as well as anything in “the DALL-E ‘house style’… or in other similar styles that humans would have trouble replicating.” In other words, he’s asking his subjects to determine authorship of the A.I. images that most resemble human art, and the human art that most resembles A.I. images.
And, of course, we’re not technically comparing these A.I. images against “human art,” but against (in most instances) JPEGs of photographs of paintings. Not to get too undergrad about it but the materiality of painting is not some accident of its being; its form, its texture, its size, etc. all carry with them meaning and effect.
And Lincoln Michael sums up the actual meaning of these results:
Imagine that I find a group of people who are unfamiliar with the works of Leo Tolstoy. Perhaps they’ve read no works of Russian literature or indeed much 19th-century literature at all. I ask them to read War and Peace by Leo Tolstoy and Peace and War by Lincoln Michel, which is Tolstoy’s novel lightly revised with some self-help bromides sprinkled in. I ask them which they prefer and 50% choose Tolstoy and 50% choose Michel. I have now proved that I am as great and creative a novelist as Leo Tolstoy. Next I give them a just dogshit Russian novel by a random writer. 60% prefer Peace and War. Now I’m as good as Russian literature writ large. QED.
What strikes me about all of this is the pettiness toward human creativity that a lot of these advocates seem to feel. From Max Read again:
“What does this tell us about AI?” Alexander writes. “Seems like they’re1 good at art.” On Twitter, people are making even stronger claims: “Scott Alexander's simple ‘AI Art Turing Test’ proves AI is creative,” says podcaster Liron Shapira. “people who categorically dislike AI art are literally wrong,” says A.I. researcher David Dalrymple.
These comments feel very much like the idea spread by the people at OpenAI that some creative jobs should never even existed. The leap from “imitative AI imitates pretty” well to “imitative AI is as creative as humans” reeks of a kind of deep disdain for human creativity. It has been a problem from the start — this lack of regard for the very people that literally made imitative AI possible. Without the blood, sweat, and tears of actual artists, the people who wrote these systems would have nothing to train their data on.
Irony is, always, undefeated.
A lot of this disdain is performative, of course. These systems are planet destroying money hogs, and so everyone that has jumped on the hype train needs something to earn their money back. Unfortunately for them, imitative AI is not really good enough at anything to really transform any part of the business landscape. The purveyors of imitative AI still hold out hope that art, an area that is almost entirely union free and not especially remunerative so there is less ability to fight back, can be that transformative space.
But I almost wonder if it is deeper than that. Elon Musk, for example, failed to give the artist credit for a piece of art of he posted and then said that giving artists credit was bad for the medium. That is the action of someone who wants to accrue credit to themselves for someone else’s work. Or note that recently an AI conference banned papers written by AI. When push comes to shove, AI people certainly seem to understand that their output is not creative, not original.
Imitative AI is just that — imitative. It has a massive plagiarism problem that likely cannot be fixed. Using an imitative AI program is not creation in any meaningful sense. If you learn to play the piano by imitating Chopin, that does not make you a great composer. To quote myself:
The prompts are not doing creative work. The prompter is not making a meaningful contribution to the process. No matter how detailed the prompt, the machine is making the composition decisions. It calculates (yes, calculates, not decides) the ultimate positioning of the figures, the ultimate lighting, the ultimate colors relationships, the ultimate other hundred other details that make a picture a picture. Arguing that the prompt meaningfully contributes to the creations is like arguing that a bride should have copywrite of a photograph because she posed the bridesmaids on the steps of the church. It was still the photographer who made the photograph through the process of all of those other decisions that go into taking a picture.
I suspect that a lot of these people only respect technological creativity. Which is deeply sad. One of the things that makes programming so rewarding is the amount of creativity involved. It is massively rewarding to have a problem, a programing language, a limited range of systems you can build on, and to wrangle all that into a solution using just your tools and your imagination. The process of creating a program and the process of writing a song are siblings. It is sad to see that lack of understanding drive so many people.
I really think these people, those who think a plagiarism machine is being creative, should just learn to fucking draw already. Or play the piano. Or do origami. Or write. Will you produce something good right away? Of course not. but that’s not the point. Creating is rewarding in and of itself. Hell, I have an entire section of this newsletter dedicated to me being a shitty writer and how much I enjoy the process of writing anyway. Using an imitative AI tool might be faster, but it would not have anywhere near the rewards that the novels and short stories I have written provide me. Creativity is its own reward, true. But the reward is not in the output, not in the being done. The reward is in the messy doing.
If you are someone who thinks that you are being creative by prompting an imitative AI, I urge you to try to do for yourself what you want the machine to do for you. It will be a thousand times more frustrating and a billion times more rewarding, I promise you. To quote one of my favorite artists “If you’re all about the destination, then take a fucking flight. The rest of us are going nowhere but we’re seeing all the sights.”
Come see the sights with us. I think you’ll be happier.