Once, several aeons and at least five years ago, I incorrectly surmised that the financial irrelevancy of my chosen career as a writer rendered me safe from the crunch of AI and Big Tech’s Terminator feet. But alas, the machines have finally come to us—the WiFi cafe stealers, the latte sippers, the people who have armed themselves in this capitalist world with nothing but their “interesting and well-rendered insights.”
Every morning, my phone screen lights up to more ill omens. OpenAI has exposed the rot at the core of the liberal-arts-degree factory; now, professors turn to AI tools to grade papers written by ChatGPT. A quick search of “how to use AI to write a novel” suggests several whimsically named tools and apps (Squibler! Sudowrite!). Novelists can now interview ChatGPT about just how soon the tool might be able to replace them, and ChatGPT can give flattering, dexterous answers. Writers (aspiring and not) could give AI models the following prompt and surely, it will spit something out: ChatGPT, could you suggest a good framework for a romantic early-20th-century spies-in-love novel, and could you make it combine the styles of Hilary Mantel, Patricia Highsmith, and Sally Rooney?
The sun of the machine apocalypse has risen on the act of writing. All remaining scribblers should lay down their squibs and come out, squinting, arms aloft in surrender.
Well, I’m here to say that we can’t let machines write our novels. Why? Because unless someone told us we were reading a novel penned by AI, we might never know the difference.
Read More: AI and the Rise of Mediocrity
This might seem paradoxical, or provocative, but I mean both halves with my whole heart. Writers and readers need each other, even though we are historically terrible at sniffing one another out. We often exist on two opposite sides of a long, dark tunnel, straining to hear whispers from the other side: Are you out there?
This connection is already so fragile, so prone to corruption. Research has shown that so-called AI detection software has a dismal success rate: As New York magazine recently reported, the program ZeroGPT identified a chunk of the Book Of Genesis “93% AI-generated.” And while I don’t think great novels will come out of a LLM, now or at any point in the future, I’m not certain about “greatness” as a safe metric to crouch behind. As much as I’d love to feel smug—a-ha! Game, Set, Match, call me when an LLM writes The Last Samurai or What Belongs To You—it seems smart to ask: Doesn’t insisting on “greatness” as the sole determinant of humanity vs machinery already admit a fairly sizable swath of lost territory? How many great novels come out of humans, in any given decade?
I ask these questions because, well, I’ve lived them. I’ve spent a long time—mortifyingly long—sweating over my sentences, hundreds of hours of my life that I could have been playing with my kid, or doing sit-ups, or attending to my finances, or cleaning under the couch, or simply standing outside, staring up into the sun. If I didn’t hope that an appreciative reader would be unable to miss the scent of my sweat equity—my humanity—in my writing, I never would have toiled so obsessively, or for so long.
And yet a smaller, colder part of me admits that a not-insignificant percentage of the work of novel writing hews perilously close to the territory now occupied by LLMs. Mapping narrative possibilities on a grid, thinking about how other novels generate and resolve tension, overtly or subconsciously mimicking the prose rhythms of certain writers—no matter what kind of novel you’re writing, really, you’re going to spend more time doing these things than you’d probably admit in a theoretical NPR interview with Terry Gross.
As writers, we can always retreat to more philosophical environs. What about intent? An AI can’t “intend” things the way humans do. An AI can craft a story, sure, maybe even a pretty good one. But it can’t want to craft a story, or obsess about how people will receive it, or stare at the bedroom ceiling, remembering an interaction they had in fourth grade that they decide has to go into the book.
As a writer, I find all of these thoughts true. They are comforting and reaffirming. I retreat to them often. However, the ground beneath here quickly grows spongy, too.
After all, how adept have we, as humans, proven to be at the act of correctly sniffing out “intent” in a piece of art?
What does the writer say, 99 times out of 100, at a public forum when a reader confidently stands to offer their interpretation for what they must have been thinking as they worked? How many times is the answer, “Wow, you’ve nailed it completely?” At best, their response will be a diplomatic “I never thought about it that way.”
These concerns are a whole lot older than WiFi. The questions at the heart of the debate about AI and writing are as old as readers and writers themselves. There is a cutting scene in Madame Bovary, for instance, in which the wealthy cad Rodolphe writes a faux-anguished “goodbye” letter to cut things off with Emma Bovary, who he has strung along for his amusement, hoping to let her down easy. He thinks of her with distant pity, even a little contempt, as he writes this missive, and then, for good measure, he dips his fingers in a pot of water and leaves a few droplets, suggesting tears, on the page.
None of us wants to be the unwitting recipient of the Madame Bovary letter. We truly want to believe that the stories we read are telling us something about the people who wrote them, and thus about people. This is the kind of collective experience that we cannot allow AI to take from us.
I have met people in the world who have read something I’ve written. I’ve introduced myself to writers I admire. And I’m here to report that once you have met with a person in the world because of something one of you wrote that the other read, you will never look at human interaction the same way again. There is something peculiar and exalted in the moment. It is awkward, but sweet beyond description. The readers’ eyes ask: Was it you who sent those messages? And the writer’s astonished eyes answer: yes, that was me. When readers and writers meet, they become, in some spiritual sense, old friends. You will be able to ask each other the kinds of questions, eagerly, that normally ruin social gatherings.
Writing takes too long. It eats up days, hours, weekends. It gives you a headache sometimes, a backache always. It is neither convenient nor remunerative work. And yet, the written word remains the only effective method of telepathy that’s ever been invented, as imperfect as it is.
We spend nearly every moment of our consciousness pawing alone in the dark, snuffling for traces of others. Removing human beings from the other end of this storytelling tunnel would double the amount of time we spend in the blackness. It would encase us in the midnight of our aloneness.
When I imagine following the voice at the end of the tunnel and not finding a human, I am crushed by a sadness too enormous to contemplate. Writing and reading are ways of believing other people exist when you are alone. If we remove that connection, we might not initially even feel a thing. But it will be one more way for us not to know each other.
The post If We Let AI Tell Our Stories, We’ll Be Lost In The Dark appeared first on TIME.
