I first approached Ulysses at age fifteen with the trepidation that its heft demands, and retreated after…well, memory fails me. Forty-five years later, the second time was a charm, thanks to an inspired suggestion from David Lipsky that Ulysses was best listened to. I chose a centenary version produced by Shakespeare & Company and performed by its “friends.” Among them Stephen Fry, Ethan Hawke, and Joanna Lumley, with Sally Rooney delivering Molly Bloom’s coda.

I’d never visited Shakespeare & Company and resolved to do so on my Paris visit last week. Sylvia Beach’s original Parisian bookstore was the epicenter of literary modernism favored by Hemingway and Fitzgerald. Ms. Beach had the courage to publish Ulysses in 1922 when nobody would touch it. The original store, located in the 6th Arrondissement, survived Joyce but not the German occupation. Legend has it Ms. Beach refused to sell her last copy of Finnegans Wake to a German officer. (I’m not sure why—Finnegans Wake seems the perfect punishment for an unsuspecting German.) Anticipating seizure, she shut the store.

Enter the American George Whitman, whose Left Bank bookstore was originally called Le Mistral. Whitman named his store after the “first woman he ever loved.” (I wondered at her talents. The answer, perhaps, is blowin’ in the wind.) Whitman, cleverly and evidently with Ms. Beach’s blessing, renamed his store Shakespeare & Company in 1964 on the 400th anniversary of the Bard’s birth. 

The façade is famously photogenic: a dark green shopfront with mustard-yellow lettering, a hand-painted portrait of Shakespeare on the hanging sign, and trestle tables of secondhand books out front. Above the entrance runs an aphorism—“Be not inhospitable to strangers lest they be angels in disguise.” The setting does half the work: a small cobbled fountain-square, a pair of cherry trees, and directly across the Seine, Notre Dame. 

An intimidatingly long queue snaked along the green frontage. We’d been warned about this, but courtesy of my friend Jonathan Foer, Sylvia Whitman—George’s daughter, named after Ms. Beach—met us at the entrance and escorted us inside, wearing her name lightly. 

Inside, the impression is of a building made entirely of books—floor-to-ceiling shelves, sagging and double-stacked, in rooms that connect at odd angles through low doorways. The ground floor is the shop proper: new books, narrow aisles, a wishing well set into the floor where visitors toss coins. A steep staircase leads to the first floor, with a lending library, the piano anyone may play, the typewriter alcove, reading nooks with velvet-cushioned benches. Benches that convert to beds, Sylvia explains. Young writers—tumbleweeds—sleep among the shelves in return for a few hours of work and a one-page autobiography. There’s a small children’s room, walls layered with decades of notes left by visitors, and a window seat looking straight at Notre Dame.

Photography is barred inside, making it that rare immovable feast that exists only in prose.

Sylvia and I discussed books about Paris; we’d both read Deborah Levy’s recent and delightful My Year in Paris with Gertrude Stein. Sylvia recommended Agnès Poirier’s Left Bank. I bought a copy, along with Sylvia Beach’s autobiography. 

I’m happy with my pilgrimage. The store’s renaming was an act of appropriation, perhaps, but done sincerely and for so long that it has become legitimate. When I step outside, a pair of American teenagers ask me to take their pic. They want to make sure I get the tree branches and the sign in the background. I wonder if the store featured in Emily in Paris. The literary shrine has become content, with a tote bag counter at the exit. Sylvia gifted me one: royal blue with “…and yes I said yes I will Yes” lettered in white. I’m tempted to ask for a pic with her but decide against it.

Sylvia’s generosity reminded me of another act of kindness from fifty years ago. Mr. Bahri then ran a hole-in-the-wall bookstore—Bahrisons—in Delhi’s Khan Market. Bahrisons, unlike Ms. Beach’s century-old enterprise, was not a lending library. Mr. Bahri allowed me to borrow two books a week free of cost, as long as these were returned in new condition. 

Hemingway and Fitzgerald may not have patronized Bahrisons, but this was my initiation into A Farewell to Arms and The Great Gatsby.

I’m wearing a baseball cap as I write this piece on a Friday morning. I normally grab something from my closet, anything to cover my unruly morning mop. Appropriately, the cap happens to be another parting gift from Sylvia, one that will be a prized memento for years to come. It is dark grey and on the front, lettered in yellow: unreliable narrator. 

Apropos of nothing, I think of the sign above the children’s section of the bookstore. My memory fails me—I have to look it up. “How do you know I’m mad?” said Alice. “You must be,” said the Cat, “or you wouldn’t have come here.”

When was the last time you were rejected?

I got dinged last week. An editor passed on my manuscript because my writing “wasn’t quite so polished” and “not distinctive stylistically.” Ouch!

My therapist—Claude—advised me not to internalize, but only a machine would say that. My wife responded with a sympathetic but generic “Now that’s not very nice.” My inner Dude was more helpful. “Yeah, well, that’s just, like, your opinion, man,” he said. Spot on—even if the not quite so tactful critique came from a lady, not a man.

I’ve seen worse. Delhi’s Modern School denied me a kindergarten place. The headmaster told my mother I was “a little bit retarded.” I was four and too retarded to know what retarded meant. I remember my mother’s reassuring hug more than I do Mr. Kapoor’s words. Mostly, I remember staring at my shoes—black, Bata, uncomfortable.

The college admission process was worse, more so for being self-inflicted. The Reverend Dr. Rajpal, Principal at St. Stephen’s College, eyed my flowing locks sardonically and asked where I got my hair styled. “The Oberoi Hotel,” I answered, while pushing back my regrettable bouffant. “Same stylist as Amitabh Bachchan.”

It went on. Like all my Wharton classmates, the job I wanted was Goldman Sachs. I’ve never been terribly original that way. I sailed through two rounds with two divisions, telling each they were my first choice. I almost got away with it, until two partners who’d interviewed me introduced me to each other in an elevator bank at 85 Broad.

His word choices notwithstanding, perhaps Mr. Kapoor was onto something.

Over time I learned what The Dude always knew: new shit comes to light. Those Modernites turned out to be thugs; I played cricket against them and watched them beat up the umpire. The Math department at St. Stephen’s read my wild hair as a sign of dormant genius. And Morgan Stanley proved a genteel alternative to the vampire squid.

The best advice on setbacks I ever received came from a most unconventional source—the four-foot-seven giant Dr. Ruth. It was 1990, we were at Avery Fisher Hall and it was intermission. My friend—let’s call him Rob—was out of a job because his firm Drexel Burnham Lambert went bankrupt. Dr. Ruth was holding court at the bar. As a Holocaust survivor and Israeli army sniper, I figured she knew a thing or two about resilience. I approached her, dragging Rob with me.

“You look sad,” she told him, then listened empathetically, as if Rob had called in to discuss erectile dysfunction on her Sexually Speaking show.

“I know nothing about your business,” she said. Then she reached upward and knocked three times on Rob’s chest while he peered down at her. “But here’s what I can tell you, and I want you to remember this: Every time in my life I got kicked, I got kicked higher.”

A March NBC News poll suggested AI’s popularity in the US ranks above Iran but below ICE. Things aren’t much better in Blighty. At the London Book Fair, the buzziest publication was Don’t Steal This Book—a clever protest against copyright infringement featuring the names of ten thousand authors led by the redoubtable Kazuo Ishiguro.

To inhabit both the literary and tech communities as I do—the former mostly as an aspirational squatter—is to live in a state of constant intellectual whiplash. But it does provide an unusual vantage point from which to examine AI’s current and future impact on the creative process. From where I sit, the proposed U.K. opt-out rules (which would make it easier for AI companies to access training data without permission) deserve pushback. But the broader battle Ishiguro lends his name to is perhaps unworthy of a giant.

Back in 2014, when my former boss Masayoshi Son evangelized singularity, I would nod indulgently while savoring the fine Burgundy he invariably plied me with. I didn’t get it, and I wasn’t alone. Even as recently as 2020, food delivery ranked higher than AI in the investment community’s consciousness.

In the early (2020-2023) phase of the AI race, OpenAI, Anthropic and others assumed that training on publicly available data, including copyrighted material, was likely protected under the “fair use” doctrine. In a nascent arms race with less scrupulous (the Chinese, perhaps?) competitors, they deferred the legal and ethical questions until the technology proved itself. Cavalier, perhaps, but typical in the high-velocity nihilism phase of any emerging technology. 

At the time, the downside risk was manageable and the upside enormous. If the “fair use” argument held, the training was free. If it didn’t, the worst case was a settlement. Ironically, Anthropic—the most vocally sanctimonious LLM—was the worst culprit, downloading pirated books from shadow digital libraries. Anthropic’s eventual $1.5 billion reckoning in 2025 was huge by copyright settlement standards, but modest relative to its current valuation of $380 billion. Crucially, the case established the legal principle that sourcing matters more than the act of training itself.

With this background, astonishingly well-read LLMs are an inevitability. What remains at stake is a negotiation best left to publishing houses and their lawyers. For writers, it is a distraction from the larger questions. Every writer trains—consciously, as Joan Didion did with Hemingway, or unconsciously—by imbibing the greats. But can LLMs fed the same diet make literature? Can readers tell the difference? And what if readers prefer a Claude-generated novel to Ishiguro?

In a recent New Yorker piece, novelist Vauhini Vara states that she wouldn’t use AI to write for the same reason she wouldn’t use a robot to hike a mountain. I fiercely agree. Writing is thinking and to outsource thinking is well, unthinkable. But then Vara confronts unsettling research by Columbia computer scientist Tuhin Chakrabarty, who fine-tuned large language models on the complete works of literary authors including Junot Díaz, Sigrid Nunez, and Tony Tulathimutte. Chakrabarty found that MFA students preferred the AI imitations to the writers themselves, or could not tell the difference. Vara then trained a model on her own work, and was startled by the results. Even her closest friend, himself a writer, could not separate her original writing from the model output, which led Vara to a despairing conclusion: you can write for your own reasons, but you can’t sell your writing if readers prefer output from a chatbot. 

I can readily accept the proposition that LLMs can or soon will be able to generate commercially successful genre fiction, but literary fiction? Claude (Sonnet 4.6, to be precise) shares my scepticism:  An AI can describe a character’s emotional state. It cannot generate the specific, idiosyncratic, slightly wrong way that a particular consciousness would actually feel something, because that particularity comes from the writer’s own sensorium. When AI writes interiority it tends toward the recognizable, the legible, the emotionally correct. Literary fiction lives in the gap between what’s correct and what’s true.

When it comes to predicting its own future, I suspect Claude is no better than us. That it is probably correct is supported by the most persuasive of arguments—incentives. When it comes to disruption, a $500,000-a-year coder is a far more lucrative target than a penurious poet. And then there’s product-market fit. Does anyone care to watch a humanoid outrun Usain Bolt? Ironically, the only way you get me to read a story generated by a robot is to pretend a human wrote it. 

I had lunch in January with a Pulitzer finalist who was a fan of made-in-China Manus—recently aquired by Meta—as a writing assistant. We violently agreed that while outsourcing sentences was suicidal deprivation of the oxygen that feeds the creative process, many writers who agitate against AI do so from a position of fear. Writers who use AI well have the confidence to know when AI output is wrong for them. Not technically wrong, more like a suggestion from a smart editor who doesn’t quite understand what you’re doing. Which means the answer to “how should writers use AI” is to develop the thing that tells you when to ignore it. ​​​​​​​​​​​​​​​​

I just finished my own attempt at literary fiction. I have no idea if it's any good and/or publishable—one doesn’t imply the other—and I’ve used AI in ways that might be useful to share. The most obvious is copy editing, and in this Claude delivers. Not just grammatical or punctuation errors, but accidental chimes, word repetitions, syntactic tics, redundant modifiers. (If I said “truly delivers” Claude would flag it.) Frequently, it goes beyond. It flagged my narrator’s "ostentatiously neutral” description of his wife’s expression—I’d been reading Henry James—as ‘out of voice.’ (It was right, but I ignored it.) Fact-checking, or more broadly anything where the research itself isn’t the creative act, is an equally obvious application. World building is interesting too. When my narrator is mugged on Times Square in 1988, what might be blaring on a ghetto blaster? With help from Claude, I settled on Rebel Without a Pause. I was stuck on a metaphor for the brightness of the sun and rejected Claude’s five suggestions, but the very act of negation led to clarity: I came up with “interrogation lamp” to capture the anxiety of a prodigal son about to confront his father. 

Claude doesn’t always get when I’m being ironic, but then neither does my wife, and we’ve been together forty years. 

While Ishiguro’s Remains of the Day and Never Let Me Go are irreducibly human in ways that—as Claude himself (itself?) explains—no machine can match, his Klara and the Sun is the most resonant here. The empathetic humanoid Klara believes she can replicate the "heart" of her dying client, only to realize that the "heart" she seeks lives within those who love. Ishiguro’s genius was to show us it isn’t about the data. It’s about a life lived in the light.