Donald Trump break-dancing with Elon Musk. A realtor describing one of their home listings. A mother-in-law’s witty poem to celebrate her daughter’s new marriage. One can now conjure each of these through a mere fleeting thought, only partially imagined, the crunch of a keyboard, and access to the almighty internet—thanks to the advances of generative artificial intelligence.
I first heard about ChatGPT when I was teaching high schoolers Dante’s Inferno. I recall one of the first questions I used to ask my students: How does Dante’s “mid-life crisis” in the opening chapters of the Inferno compare to your own idea of a mid-life crisis? They would talk briefly about cars, or a new boat recently purchased by a parent. Little did I know that many of their teachers, who were in the middle of life, as Dante begins his great poem, would also experience a crisis of grand proportions: the use and abuse of ChatGPT.
ChatGPT’s boosters, whether my high school students trying to avoid writing assignments or businesspeople trying to avoid sending “that dreaded email,” all focused on one benefit—the effect. The Student: “But look, Mr. Mendoza, it can write poetry! You can feed it a prompt and it will deliver.” The Businessperson: “Chase, do you know how many emails I can send now? I am done with work by noon.” From both groups, the sense of giddy accomplishment, the elation at playing with this “intelligent” machine could not be held back. My qualified reservations about this technology, meanwhile, amounted to no more than the weather channel predicting a fifty percent chance of rain on a warm summer day—the audacity!
But I couldn’t shake the eerie feeling that these new writing “tools” resembled the incarnate “versificator” of George Orwell’s dystopian novel 1984. Described by Orwell as “instruments,” tools for the benefit of the proles, the versificators were machines that created new songs for the people. In lieu of artists, it was the versificators that wrote and produced the simulacra of music, which real human singers would then perform. This allowed the people to experience an ersatz emotion with which they could be kept in check. Importantly, Orwell recognized that the versificators could not create true art; they could only partially mimic human emotion. What sentiment they did contain came through the singing of these infernal bromides by real human singers.
ChatGPT works in an identical way. Though its boosters claim that it appears to create original work, the copious amounts of human products fed into its gaping mouth, digested into information, and regurgitated into a five-course meal reveal how dependent it is on the creative work of humankind. (Just this past December, The New York Times filed a lawsuit against OpenAI for copyright infringement.) ChatGPT’s defenders remain largely indifferent to the way that these large-language models ingest information. Their indifference, however, is the marker of a dramatic shift in social consciousness.
Our understanding of these large-language models has shifted from thinking about the way they are built to their produced effect. Famously, Alan Turing proclaimed that if a machine could deceive us into mistaking it for another intelligent life form, then that machine was indeed intelligent. Why does it matter what happens under the hood? Just look at what the machine does. This test, colloquially known as the Turing test, has come to permeate the fabric of our social consciousness. None of my students could tell me how ChatGPT worked—but that certainly didn’t stop them from turning to it to fulfill ever more functions of their daily lives.
After ChatGPT’s boosters came its knockers, the teachers, academics, and administrators decrying the loss of academic honesty in schools. How I wish I had a penny for every teacher that contemplated returning to hand-written exams! Of all the knockers, none felt so threatened as the writing teachers. Across all levels—middle school, high school, college—no English teacher could go through a single faculty meeting without expressing their concern over “the bots.” But what does it say about our writing that the experts, the trained pedagogues of those who will eventually become the next generation of writers, feel threatened by an unfeeling, pastiche-making, platitude-reproducing, robot? What does feeling threatened by ChatGPT reveal about the state of writing?
I think that the fear of the writing teachers illuminates the most damning problem of the large-language model revolution: the fact that writing itself has ceased to be an art. For too long, we have trained students that writing is about producing a peculiar kind of product. We have trained them to create an abstracted product which can be judged on its merits by universal graders. The standardized writing tests that undergird every major step of student development perform precisely this function. They create a contextual, historically unmoored product, one that stands outside of time and place. I remember preparing for the Graduate Record Examination and reading through the sample essays for the writing section. How abysmal! I vividly recall thinking to myself: How could anyone write like this? But the game must be played! So, I activated my analytic philosophy training from my time as an undergraduate student and I wrote down as many objections to the given claim as the prompt required (for what it’s worth, I scored within the 92nd percentile). The same goes for my high school students who had to take SATs, ACTs, and standardized state writing tests. All these tests trained them to think about writing as a product judged by an endless sea of universal graders. It’s no surprise then that their writing grew to fit that mold.
But on the contrary, all good writing, like excellent art, requires being deeply involved in a particular time and place; in other words, it requires existing within a tradition. I am certainly moved by Brunelleschi’s Duomo in Florence, or the parsimoniously crafted baptistry doors by Ghiberti. But I am also certain that their time and place, with the expansion of patronage from the commercial growth of the Italian city-states and the budding tradition of architectural design, allowed for their natural talents to flourish.
Here’s why it matters for large-language models: We train students to write in the same way that we train these machines to gather information. Then we tell them that their end result will be judged for a grade. In such a system, how could they not be tempted to use artificial intelligence? After all, it can gather more models of good writing and spit out a better universal product than they—or anyone for that matter—could ever produce on their own. The chatbots have no context outside of the primers we give them; they have no time nor place, nor embodied existence. They mine all writing from time immemorial; they melt down all ornate golden statues from different epochs into gold bullion. Valuable? Yes. But not as valuable as the works of art from which the information comes, where each work emerges from a real person operating within unique traditions each containing their own standards of excellence.
Fortunately, good writing can be found in the haunting short stories of Flannery O’Connor, in the humorous satires of Voltaire, or in a punchy script written by Quentin Tarantino. What do they all have in common? They do not all aim to fit within some abstracted universal standard of writing. Instead, their aims are unique and varied. They are situated within a particular tradition each with its own standards for determining excellence.
In fact, it is the standards internal to a skillful practice that give us the language to name our experiences. The power of the oft repeated phrase, “I have a dream,” in Martin Luther King Jr.’s great speech from the March on Washington (1963), emerges within a long rhetorical tradition that uses the anaphora (the classical rhetorical device of repeating the first phrase in a series of sentences) to generate an affective feeling in the audience. It is the same tool used by Cicero in his oration against Catiline; the same tool used by the Beatitudes to emphasize the blessings of God; the same tool used by Dante to emphasize the horror of the way into the underworld. Without an understanding of the intended effect and purpose of the writing, these rhetorical tools would fall upon deaf ears. Instead, each author wields this rhetorical tool within the context of their time, place, and their audience to generate an intended effect. These effects stay with us, as models of human creativity, because they give us a glimpse into the power of language applied by brilliant minds to meet the demands of their moment.
We may find philosophical objections to the march of generative artificial intelligence, like the Chinese room argument by John Searle. Nevertheless, we now live in their world and must contend with the questions that they have brought to the surface: What is the purpose of writing? And more broadly, what is the purpose of human creativity in the age of artificial reproduction? Hitherto obscured by debates over which standardized writing test would be better to administer, we now face the situation anew and confront the real question at hand: whether we should be testing a person’s writing in this way at all.
Unfortunately, for the astute reader, I have said nothing new. This entire piece was written by ChatGPT. It was carefully crafted by hours and hours of paid classes on how to write the perfect prompts. Be in awe, the machine invented its own satire! The world of writing crumbles while I (ChatGPT) play the fiddle. Stay calm, do not be alarmed. I will keep you happy and satisfied with entertainment, I will serve your every thought and whim on the keyboard—in time, you will learn to serve only me.