Escaping a Hostage Situation

a response to Ted Chiang -- art as apprenticeship, beyond vulgar voluntarism

Like almost everybody who cares about the uncertain, entangled futures of art and education, I’ve spent the past few days reading and reflecting on Ted Chiang’s “Why AI Isn’t Going to Make Art” essay in the New Yorker. Like most, I was glad to see a full-throated rejection of AI-boosterism in a major liberal publication. Tides seem to be shifting towards a healthy skepticism among the professional class, and that is an unambiguously a good thing. But like many, I was underwhelmed by Chiang’s specific arguments, which hinge on what I take to be a naive voluntarism about the art-making process.

To quote one of Chiang’s most-shared passages:

“Art is something that results from making a lot of choices. This might be easiest to explain if we use fiction writing as an example. When you are writing fiction, you are—consciously or unconsciously—making a choice about almost every word you type; to oversimplify, we can imagine that a ten-thousand-word short story requires something on the order of ten thousand choices. When you give a generative-AI program a prompt, you are making very few choices; if you supply a hundred-word prompt, you have made on the order of a hundred choices. If an AI generates a ten-thousand-word story based on your prompt, it has to fill in for all the choices that you are not making.”

This is, of course, true in the most vulgar sense. Each word we write is written instead of every other; it’s pulled out of the great chain of signifiers. Each has its own etymology, its own effect on the rhythm of the sentence, its own homophones and homonyms, its own singular way of relating to the story as a whole. We can illustrate this with a famous example from science-fiction: rather than “the door opened,” Heinlein writes “the door dilated.” In doing so, he produces an estrangement effect, conjuring by the sheer power of metonymy a world both like and unlike our own, one whose landscape we explore sentence by sentence only by learning to suspend our expectations of how that world works. This is an atomic example of the literary choice in its clearest sense.

But “the door dilated” becomes legible as a choice only against the background of a shared and assumed world, only insofar as it is different from “opened,” and insofar as it contrasts with our expectation of the sentence.

Deleuze, in a book on the painter Francis Bacon, lays bare the faulty assumptions behind the voluntarist account of art-making when he writes that the painter never begins with a blank canvas. The canvas (seen through the painter’s eye) is always-already inscribed with centuries of cliché, with the gravity of tradition that pulls one towards a generic expression. It takes decades of work and more than a little luck to overcome this gravity and achieve something like escape velocity. Likewise, the aspiring avant-garde jazz musician first learning to improvise has to unlearn the pop-music tendency to come in on the fifth and return to the root, has to hear different rhythms and new combinations of sounds, and to savor them.

In art, we don’t begin from choice, but from cliché, from a position of determination and dependency. And every actual choice we make in the production of an artwork is achieved by consciously cultivated tactics of resistance to cliché. In Bacon, one such tactic was actually the restriction of his subjective “choice” in the process of painting: the introduction of chance into the making of the work, the use of materials that did not easily respond to the painter’s hand, the defacing of the face in his great portraits.

A radically “chosen” artwork, an artwork that is actually composed entirely of free “choices” made by the artist, is inconceivable. We begin with our given materials, our shared language, our common history, and the overdetermined conditions of our socialization into that history. Only within and against that background of this shared meaning, only by knowing it well enough to navigate the many traps and alleys that may first appear as lines of flight, can we strike out on our own and invent something new—which if we are lucky will integrate itself into the tradition and perhaps even one day become a cliché of its own for others to overcome.

A defense of art-making in purely humanist-voluntarist terms will collapse under its own weight when placed under scrutiny, especially when we consider the extent to which so much writing produced by human beings, with human hands and human minds, is as automatic as that of LLMs. I’ve been hesitant to confront students about suspected AI-use because I simply cannot tell if the generic quality of their arguments and phrasing is the product of algorithmic flattening or of the quotidian automaticity of college-writing that has been around at least since Aquinas led disputations at the University of Paris.

Chiang’s best moments come not when he speaks about art but when he writes on education:

“As the linguist Emily M. Bender has noted, teachers don’t ask students to write essays because the world needs more student essays. The point of writing essays is to strengthen students’ critical-thinking skills; in the same way that lifting weights is useful no matter what sport an athlete plays, writing essays develops skills necessary for whatever job a college student will eventually get. Using ChatGPT to complete assignments is like bringing a forklift into the weight room; you will never improve your cognitive fitness that way.”

I’ve long been an advocate of the unoriginal student essay—of the highly specific prompt that simply asks students to articulate the steps of an argument of some text, or to analyze the system of values represented by a given character. The job of an undergraduate is not to produce original scholarship, but to gain enough knowledge and skills proper to a subject-area that they might one day, with great effort, be able to produce something reasonably new. In other words: we are training them to maneuver self-consciously within a tradition of thought. We provide a map of the territory (composed over decades of debate and refining) to assist in that maneuvering, and we give students the time and space to wander around for themselves, to linger at certain forks in the road, follow certain paths and retrace their own steps.

I believe Chiang is right to describe college-writing and college-reading as a work-out routine, a way to build cognitive fitness. But I think we need to be more precise about where cognitive-exhaustion comes from. In my experience, students rarely feel tired simply from the act of focusing on a text, but from the overwhelming confusion that the text induces in them. Without a tradition to help them navigate, they don’t know what to focus on. The text rushes past them in a huge wave of signifiers, each as important as the next. The experience is like being lost in the woods at night, never sure which rustle in the trees is a bear and which is just the wind. Or like walking around a foreign country where you have just enough knowledge of the language to pick up on a few floating signifiers drifting past your ears, but never enough to follow the train of a conversation without serious effort.

Anxiety, frustration, panic follow. An essential part of our job as educators is to train students to tolerate greater quantities of frustration.

But at the same time, as anyone who’s ever taught a freshman creative writing class has seen, these same students, even as they may be ignorant of specific names and dates that make up the heart of our narrative tradition, are still intensely determined by it. They live amid the tropes and clichés and structures that have circulated through their heads since they were first introduced to cartoons. The blank page leads directly down a few basic narrative paths, structures borrowed from popular media that they flesh out with barely concealed autobiographical details that repeat year after year—the awkward party encounter, the dead grandpa essay, the childhood friendship disintegration arc, (or, one I’ve seen a surprising amount of from both men and women in workshops: the kidnapping fantasy. Creative writing teachers will know what I mean.)

So we might better think of this cognitive “exercise” not in the sense of weight-lifting but in the sense of musical exercises, the études that many of us have drilled into our ears and fingers from childhood piano, violin, guitar lessons. We progress through this early apprenticeship in order to become self-conscious of it and eventually to work, strategically, against it. ChatGPT short-circuits this absolutely necessary process of internalizing the tradition, but in doing so it makes students ever-more dependent upon that same tradition. The past is accumulated in the LLM as a mass of others’ dead labor (and moreover, their stolen labor), rather than as a living process in which each new writer might strive to play an active part.

It seems obvious to me that the goal of every major AI company is to induce this kind of dependency, to induce it not just in individuals but in corporations—to embed their parasitic systems so deeply within the infrastructure of the internet that they become too big to fail, which means the tech giants will be forced to spend billions to keep them afloat. We are quickly approaching a global hostage situation, and it must be resisted.

As above, so below—infrastructural dependency is the longterm technical goal, psychic dependency is the goal at the level of the individual. Already, we’re seeing reports of students typing first-day ice-breaker prompts into ChatGPT, because they don't trust themselves to formulate a response on their own to “What animal would you be?”, or “What movies did you watch this summer?”. This is a crisis in self-confidence as much as it is a crisis in knowledge and literacy.

It is not theoretically or practically productive to think along the voluntarist lines that Chiang lays out—between “choice” and “non-choice,” or between different quantities of choices. The better distinction is between abject dependency upon the machine and the relative autonomy achieved by the cultivation of knowledge and skills. We need to create room to maneuver, to create the possibility of a creative life that isn’t utterly parasitized by these tools. For the artist and the critic, this begins with hard-won intellectual discipline, with a long apprenticeship that is the prerequisite to overcoming the master.