For much of the past year, I’ve been torn on an issue that matters to almost no one. But maybe, just maybe, you’re as nerdy as me, as addicted to abstract literary arguments that no Ph.D. would touch, as willing to pursue a thought to its absurd end. If so, then here’s the conundrum that’s been keeping me up at night:

First: AI writing is bad. This has been much discussed! The New York Times has been all over it, from a lengthy essay in December (Why Does A.I. Write Like That?) to this recent quiz, which asked readers to choose their preferred writing samples from a series of pairs across five genres, from literary fiction to science writing to poetry — one of each pair was written by a human, the other by A.I. Most readers chose the A.I.-written samples. Including me! Not because I thought they were “better” than the excerpts by, say, Cormac McCarthy or Ursula K. LeGuin but because they were simply cleaner than the human authors’ tortured, precious prose. In none of the cases did I think the writing was particularly “good,” and only in the poetry category did I “prefer” the human, Elizabeth Bishop. But, of course, we all know what I think of poetry.

This test, though, was a fake one: It reduces writing — novels, journalism, poetry — to a basket of sentences, asking us whether the barest, most arbitrarily selected part can stand in for the whole, and stand up against the output of a prediction machine that’s been trained on the corpus of all humans and asked to generate a handful of characters in imitation of its mentors. It’s a false, artificial choice. A real version of this would offer me the chance to read all of Blood Meridian plus an entire A.I.-generated novel, and only then ask which I preferred. Perhaps McCarthy’s “sparse punctuation … archaic vocabulary, and … biblical cadence” (as Google’s AI summary describes it) would finally appeal to me? But even if it didn’t, I can say for sure I would prefer it to whatever the LLM cooked up.

And that has nothing to do with the quality of the prose or the coherence of the story. It’s about intentionality. Cormac McCarthy had it. In all of his work¹, he was attempting to communicate something he felt or believed — about how the world functioned, how human beings behaved, what it all might mean or not. The words on his pages constitute a unique point of view rooted in reality and experience, and whether I like that POV or not, whether I think it’s “good” or not, that origin and McCarthy’s desire to communicate it instantly elevate it beyond anything an A.I. engine can create. A.I.’s aren’t capable of intentionality, merely of mimicking it at the behest of their prompters — often convincingly! But there’s nothing beneath the surface. It’s sophisticated bullshit, an attempt to pass off calculated guesswork as the product of earnest, lived experience, an impossibility for a computer program. And bullshit is the province of conmen.

And so but here’s my problem with this line of thinking: At the same time, I believe in the death of the author.

More after the not-an-ad ad…

“The death of the author,” as you no doubt remember, is the literary theory espoused in 1967 by Roland Barthes, who argued that readers — critics, academics, reviewers, and regular old people — had ceded too much power to authors in interpreting their work. The cultural Establishment went looking for biographical details in novels, listened attentively to writers explicating their narrative decisions, and essentially put authors on pedestals.

Barthes, however, wanted to free readers from that cult — “To give an Author to a text is to impose upon that text a stop clause, to furnish it with a final signification, to close the writing“ — and return to them the power of interpretation, experience, emotion. They (we) make the meaning of writing (or art, or music, or whatever), not its creator, who may or may not even understand the range of meanings he or she has created. (We writers can be surprisingly limited in this way!) Barthes saw this as a kind of liberation, and it really is: You can read a book without ever wondering, What is the author trying to get at here? and instead decide for yourself what the words — the text, as theoreticians would have it — amount to. As Barthes wrote, “the true locus of writing is reading.” And without readers, writers are dead anyway.

So, how do I square these two apparently conflicting ideas? Is an actual author essential for a work to have meaning? Or are actual authors, corporeal or not, beside the point when it comes to our finding meaning, pleasure, relevance in a text?

I wouldn’t say this conundrum has kept me up nights, but it’s certainly something my mind has churned again and again for months without resolution. Until a month or two ago, when I asked myself this:

What happens when an author dies?

Well, first of all, everyone is very sad. This is America, of course, where our literary heroes — especially travel writers and producers of ersatz philosophical essays — are held in the highest esteem! There will be rending of garments, wailing in the streets, live televised speeches by our most powerful politicians and public figures. Much, much sadface.

But then what happens to the author? After the lying-in-state at the Capitol, I mean. It’s what happens to all tortured, creative souls: They become a ghost. For all eternity, or at least while their works remain in print, they are condemned to haunt the earth, revisiting their triumphs and their failures, and making spooky noises in the homes of anyone who ever gave them fewer than four stars on Goodreads.com.

And so this is the metaphor I’m using: the ghost of the author. If we readers are going to kill our creators, we should expect to be pursued by their ghosts. We won’t always see them or hear them, and we won't always understand them if we do, but we want to sense them when we read and to feel the hairs on the back of our necks stand up in their otherworldly presence. I want books and movies and music and art I can interpret in my own way, but I also want to acknowledge the dearly departed² without whom these works would not exist at all. The best work has a way of conjuring up these spirits, of making you feel that viscerally ethereal connection whether you were seeking it or not.

A.I. can’t do this. What’s never lived can never die, can never haunt its realm. The words of the LLM’s dried voice remain quiet and meaningless, a whimper drowned out by the ceaseless shuffle of true ghosts.

Still, let’s be honest. The code can be convincing — it gets more so every minute. You may someday read an A.I.-crafted story that begins to persuade you, whose language and characters and form stir a sudden, spooky breeze in the still room of your soul. And you may wonder, Have the computers finally done it? Do I sense the ghost in the machine? And that’s when I would remind you of what we all also know about ghosts: They’re only real if you believe in them. Turn the lights on, and they go away. 🪨🪨🪨