by John Dalton

from Vol. 38, Nos. 3 & 4

ChatGPT, please write a haunted house story in the style of Jane Austen.

In the quiet hamlet of Willowbrook, nestled among rolling hills and ancient oaks, stood a grand manor known as Hawthorne House. It was a place of beauty, its stone façade adorned with ivy, its windows sparkling in the sunlight. Yet, despite its outward charm, Hawthorne House held a secret that sent shivers down the spines of those who dared to tread its hallowed halls. Miss Eleanor Fairchild, a young lady of delicate sensibilities, found herself unexpectedly invited to a soirée at Hawthorne House. She was a woman of fine breeding, with an affinity for novels and needlework. Her curiosity was piqued, for the whispers of the townsfolk had long painted the manor as a place of mysterious occurrences.

2.

I’m ready to go out on a limb and write blurbs for the first few literary novels produced by ChatGPT 4 (as well as versions 5 and 6).

“This came into existence astoundingly fast. Every page of it is competent but otherwise unremarkable.”

Or

“Here is a novel that superficially imitates the style of Jane Austen. It’s for readers who never understood what made Austen an exceptional writer in the first place.”

3.

It’s fun to bad-blurb ChatGPT—fun with no possibility of meaningful blowback. ChatGPT has no defensive regard for the Austen knock-off it threw together. So join me in pointing out how inadequate it is, and feel the flush of pride in what only humans can create.

Now let’s consider what’s coming our way with ChatGPT 7, 8, and 9.

4.

AI isn’t really writing. It’s compiling statements. And it’s proving itself to be an increasingly sophisticated compiler of statements—of words, phrases, and sentences. I don’t know enough about AI learning to know if it understands what it’s compiling or even if understanding is the right term to apply to AI.

I do know that, unlike human writers, who improve unevenly but steadily over the course of many years (until they don’t), AI improves exponentially. Those superficial and lightning-quick imitations of Jane Austen are amazing, even if their literary effects are unremarkable. And they won’t stay unremarkable. AI can generate thousands of such imitation novels in the time it takes me to write a short essay. Perhaps then, it can figure out which of its Austen pseudo-novels creates interesting and worthwhile literary effects, and which ones are duds. It can hone in on the worthwhile literary effects while minimizing the unremarkable passages. Through something like trial and error, it can generate a novel that has seven to fifteen genuine, Austen-like pleasures per page (the precision of Austen’s social comedy, her quotable pronouncements on human behavior, the Austen wit with its eye for moral failure). I’m saying that ChatGPT 9 may go beyond Austen stereotypes and whip up a very satisfying and sophisticated confection of Austen reading pleasures and place it in the context of a Regency-era romance or gothic fable. Or a London-set steampunk story. Of course, these plot imitations will be much easier for AI to pull off than the imitation of tone or the imitation of a sharp, self-aware, and highly discerning intelligence.

What ChatGPT 10, 11, and 12 will be able to do is harder for me to imagine.

5.

This is the point in the essay where I reassure you about the impossible-to-reproduce qualities of the human soul.

I’m not sure that’s the right way to think about what’s coming our way. Because I suspect that AI will soon be able to outwrite us—or at least out-compile us.

Perhaps the more useful thing to think about is how a super advanced, art-producing AI will influence and (unintentionally) counterinfluence our taste for popular entertainment and for art. What will AI lead us to want in a novel, a TV show, a piece of visual art? More important, what are the elements, no matter how vast an AI intelligence, that it cannot provide a human audience?

6.

Thirty-plus years ago, I had the experience of traveling alone for many days through parts of northwest China where I was a distinct outsider. People were always staring or scowling at me from the confines of their unhappy lives. I was always looking on—from a window, a passing train, from a crowded city sidewalk—while a hundred thousand unknowable beings eddied around me. It was a rich and estranging experience. And the thing that repeatedly saved me from crossing over into a state of helpless alienation was reading a long, complex novel I’d brought along. I’d sit down outside a train station, pull out this really good novel, and the voice of the novel, the calm assurance of a human intelligence imbedded with the writing, would more or less save me. I was saved, again and again.

I’d almost rather not reveal that the novel was Anna Karenina. Because it’s such a dignified, look-what-great-taste-I’ve-got sort of choice. But it was Anna Karenina, even though it could easily have been one of several hundred other less esteemed novels I’ve read over the course of my adult life that have kept me emotionally and psychologically grounded. “Fiction is one of the few experiences where loneliness can be both confronted and relieved.” This statement puts what I experienced in China into context for me. I’m very grateful for it. (It was compiled by an intelligence known as David Foster Wallace.) I would add that fiction is especially good at confronting and relieving loneliness when you’re away from home and your usual routines and out in the world traveling alone. During my traveling days, I would read a good novel and be consoled by an expansive sense that another human being had looked bravely, calmly, methodically into the multi-faceted and inexplicable thing we call “life” or “reality” and was able to convey an attitude I found reassuring. The attitude went something like this: Hey there, tender reader. Let me use my words and lived experience and my imagination to bring you as close as possible to the endlessly varied and complex thing we call being human and alive and to the surroundings we call reality. Let me skirt past the superficial tropes of what reality is, and let me bring you close to the actual thing. As close as possible. But never all the way because, my tender friend, we humans simply don’t have the words or bandwidth to really comprehend the reality we’ve been dropped into. So closer will have to do. By being closer, you’ll recognize patterns you wouldn’t have been able to notice before, and you’ll feel beneath the events and ideas of the story a kind of whispered insistence that, though reality is weird and unpredictable and lonely, it’s also quietly wonderous. And if this whispered voice is humble and refuses to offer easy consolations, we’ll know that the strange processes we all endure as living beings (of having wonderful people and wonderful abilities granted to us and then having them stripped away) is, in a loose sense of the word, survivable. We won’t literally survive reality, but we’ll get through it, somehow.

Will the novels ChatGPT 9 compiles be able to offer us this consolation?

7.

The implied answer, you may think, is no. But I’m not at all certain. I have no practical, hands-on sense of how AI compiles its statements to make the story I posted above. To the average person the difference between the superficial Jane Austen and real, self-aware, witty, discerning Jane Austen would be tricky to detect. And not because they’re stupid, but because they haven’t spent thousands of reading hours absorbing the nuances. Those nuances wouldn’t seem like much of a difference to people who don’t read fiction. But it might be an impossible chasm for AI to cross. Or, it might well be a crossable chasm, but just not worth all the expense and human work hours needed to train the AI. There are, after all, a few bigger problems facing mankind than getting us from subpar AI Jane Austen imitations to superior AI Jane Austen imitations.

8.

But I wonder if, in the range of twenty or so years, we might get an artificial intelligence so complete that it could offer us novels and paintings and music and movies equal or better than what great human artists have created. Would the future audience be put off by the fact that no human creator had struggled and suffered and persevered to bring this magnificent creation into being? I suspect that most of AI’s future audience won’t worry much about that. There will be holdouts and objectors, certainly. But the larger audience may come to see AI in different terms: as a composite of the best human artists and thinkers. They’ll have a sense that the best of Austen and Tolstoy are embedded within that AI. So, too, the storytelling skill of J.K. Rowling and Stephen King. They’ll feel the intellectual signatures of Seneca and GeorgeEliot and Hannah Arendt. It will all be a blend of human thought and creativity supercharged by AI, the same way that all the household objects we will utilize in our daily lives will be human creations supercharged by AI.

As a concept, this seems sensible. Yet the shrewder part of me knows that every buoyant possibility offered by technology also brings with it unforeseen problems and fissures in the social order. Plus, we’ve been over-promised things before. Throughout the ages, all the best human writers have known to be wary of their own assertions. Is ChatGPT able to do that?

9.

ChatGPT, in a hundred words or less, tell me if it will one day be possible for AI to write novels as intelligent, imaginative, and self-aware as the best human writers.

It's uncertain if AI will ever match human writers in creating novels with the same depth of intelligence, imagination, and self-awareness. While AI can mimic patterns and generate text, it lacks genuine consciousness and emotions. Future advancements may yield more sophisticated results, but the essence of human creativity and introspection could remain beyond AI's reach, limiting its ability to produce novels on par with the best human writers.