Although I teach and edit fiction, I still didn’t think it would ever happen—that an AI-generated or AI-assisted short story would win a major contest. Then came Jamir Nazir’s “The Serpent in the Grove,” a regional winner of the Commonwealth Short Story Prize, published on Granta magazine’s website. The story exhibits all the common AI-writing “tells” familiar to anyone who has ever scrolled through a restaurant’s Instagram feed. Em dashes, the word hums (in the first sentence), contrasts (“It’s not X, it’s Z”). No final verdict has been made about the story’s creation, and we may never know for sure whether Nazir relied on AI to write it. Yet the preoccupation with who the author might be—man? machine? cat?—distracts us from the real casualty of AI-written fiction: the reader.
Soon after “The Serpent in the Grove” appeared, one Reddit user concluded that “AI-written or human-written, it’s painful to read.” I agree: The metaphors are nonsensical— “Laughter can cut a hush, not cure it”—and the narrative is hard to track. I followed along merrily as other Reddit users performed their own close readings. Suddenly, I was back in the college classroom, or a version of it, as words and phrases were parsed not only for evidence of a hoax but also to make judgments about their quality as literature.
The conversation didn’t change my mind about the story’s literary value. But the public attention to the story’s style made me think about the tight relationship between a reader and a work of literature, and the ways that the advent of AI may damage that bond, if it hasn’t done so already.
Reading is a solitary activity, not a lonely one. I was a shy kid who loved books in part because they felt like a form of friendship extended to me by the author. The phrase Dear reader, common to fiction for centuries, may be conventional, but it also suggests a relationship between author and reader that is intimate, warm, even affectionate. The literary critic Wayne Booth, in an essay on George Eliot, writes that “all stories, in short, claim to offer something to us that will add to our lives, and they are thus like the would-be friends we meet in real life.”
Of course, not every author is pleasant company. And sometimes the connection between writer and reader becomes pathologically intense. Interest in the author’s life can curdle into obsession, as in Henry James’s The Aspern Papers, in which the narrator embraces any deception necessary to acquire the letters of a dead poet. But in its milder versions, the bond with an author leads from admiration of the author’s style to curiosity about their life. Biographies of authors consistently outsell other forms of literary criticism. When I say that I teach literature, one of the most common reactions people have is to ask me whether Shakespeare really wrote his plays. Yes, mostly, but the point is that we imagine someone, an author, a proper name, at the other end of the line.
[Read: This literary AI scandal changes everything]
Claude has a proper name, and the line is always open. As the technology improves, style will be harder to use—it has already become harder to use—as a clue to the truth about authorship. Services are available that humanize AI tics. The consistency of a literary style over time may also be replicable. If AI can be used to write a story, an AI agent could presumably be cloistered with only the text that it has generated until a corpus of works emerges on the scale of Balzac’s or Zola. Maybe it will be possible to write a literary biography of someone’s AI assistant; maybe daily conversations with Claude will be studied for the slow cultivation of a singular way of writing or speaking.
But Claude, or whatever your AI agent’s name might be, is not an author in the way that Booth describes. Because when I sit down to read “The Serpent in the Grove,” I am no one’s “dear reader.” Instead I have the cautious, paranoid, seasick feeling that, from sentence to sentence, I don’t know whether the author is my friend, Jamir Nazir, or the machine that may have helped him write. So even if reading a novel by Claude and one by Virginia Woolf are, in some narrow sense, the same activity, reading as a practice has changed. What readers might never regain is our confidence that a text has been “fixed by the nature of human life itself,” as the philosopher Stanley Cavell writes.
When I sat down to read Nazir’s story again, a month after it appeared, I was in Gambier, Ohio, the home of Kenyon College. The Kenyon Review, founded in 1939 and edited by John Crowe Ransom, published some of the greatest writers of the 20th century: George Saunders and Joyce Carol Oates, W. H. Auden and Rita Dove. As a young poet, Robert Lowell came to Kenyon to study with Ransom. On this weekend in late May, the students were gone, the Adirondack chairs were beckoning, and I couldn’t imagine a better place to sit and read.
As I opened the story in this place that felt custom-designed for making friends with authors, I thought of the heritage of a little magazine, a venue where stories are carefully selected and meticulously edited, where students like Lowell came to learn how to make decisions about placing the right words in the right order. And I thought about what AI is taking from us. As readers, we are all losing friends, and everything that comes along with friendship: trust, distraction, escape, absorption. These are the inchoate and unarticulated aspects of the relationship an author offers to us through a book, the parts of the reading experience that provide a kind of psychological mooring for a reader. The critic Roland Barthes thought that our fantasy of an author distracted us from what a literary text really was, namely, “a tissue of citations, resulting from the thousand sources of culture.” Barthes argued that we’d always find new ways to deny the “death of the author,” but AI may have written the epitaph.
[Read: AI-writing scandals are getting very confusing]
In the absence of the author, or of the certainty that one exists, we may default to a style of reading that is self-conscious, hyperaware, restless, and anxiety-driven. We may struggle to immerse ourselves in a book, and instead hover at a safe distance, or dip in and out, worried that we’ll be fooled once again. Nothing less than the pleasure of reading is at stake.
Writers write from compulsion, from necessity. That’s why prose can feel hot on the page. I think of these two perfect sentences from Faulkner’s Go Down, Moses: “No wonder the ruined woods I used to know don’t cry for retribution! he thought: The people who have destroyed it will accomplish its revenge.” Those sentences were created by a mind that blisters from its contact with reality. AI-written stories may produce prose that wins contests, but the urgency of the desire to get the words down is never going to be there. Even if it goes forever undetected, the absence of that urgency should matter to us, just as all fakery should matter to us: a fictionalized autobiography, the imitation of a Monet. In the absence of the occasion that startles the writer into action, the ersatz artifact is meaningless, no matter how closely it mimics the real thing.
I doubt that humanity will lose the ability to write urgent sentences, but I’m worried we’ll approach them with suspicious minds and shuttered hearts, unsure about what it is that we are reading. “All fictions,” Booth writes, “come not as tricks but as gifts.” That may not be true anymore.
The post The Death of the Reader appeared first on The Atlantic.
