It Was the Blurst of Times: AI Writes a Novel
It’s one of my favorite jokes, from an early episode of The Simpsons: Mr. Burns shows Homer a room where the billionaire has a thousand monkeys typing away at a thousand typewriters—each one chained to his or her machine. “Soon,” Monty promises, “they’ll have written the greatest novel known to man.” (An odd ambition for the owner of a power plant.) He snatches a page of a work in progress—the monkey has taken up smoking, naturally—and reads out the opening line: “It was the best of times, it was the… blurst of times! Stupid monkey!”
The blurst of times has arrived. Now you can ask ChatGPT or some iteration thereof to write a novel for you, based on whatever criteria you choose. A love story set in a cemetery, with horror elements? Sure. A war story set in an opera house, with vampires? You bet. A Star Wars spinoff set in your neighborhood, starring you and your friends—and your enemies? Here you go. Almost instantly.
Now, I’ve only seen it spit out several paragraph-long plot summaries of these “novels”—perfectly acceptable outlines, the sort of thing an experienced writer could flesh out into something interesting, readable, and even innovative. Presumably parameters can be adjusted to produce full-on book-length manuscripts. Or shorter, if you just want something for your forty-five minute commute. Or much longer, if massive doorstopper epics laden with side quests and scenery are your jam.
Behold, the mash-up of mad libs and fan fic—in the style of James Joyce or Colleen Hoover, should you choose.
The Patterson Model
This is nothing more than what James Patterson has set up with his shop of ghost-writers. He gives them a page, or even just a paragraph, of notes for each book—a set of parameters that you could just as easily give to an AI—and the writer returns with a finished manuscript, at an agreed-upon deadline and for an agreed-upon fee. Patterson has established himself as an easily identifiable brand, and readers trust him to deliver a product that they can be confident that they will enjoy and that will be up to a certain standard. And so he is a multimillionaire who no longer really writes, and yet is one of our more successful “writers.”
The Patterson model will be easily (and infinitely) replicable with AI. An AI bot will be able to generate, edit and proofread the manuscript, format the book for all platforms, design the cover according to current trends, and give a reading with appropriately gravelly AI-supplied “voice actors” for the audiobook. The AI will A/B test the cover and tagline to see which perform best on the algorithmically driven sales platforms. (Presumably AIs will provide reviews and blurbs as well, as sources like Publisher’s Weekly and Kirkus get swamped with submissions, and as these publications realize it’s cheaper and easier to have bots read and digest new books.)
I am confident that in the next year or two Amazon will start to be flooded with AI-generated novels, as “writers” game the system and swamp certain categories with cookie-cutter prefab novels with attractive covers and taglines, where the “writer” is essentially setting up the sort of shop that took Patterson decades of actual writing, serious effort and trial-and-error to establish. And if you think that Amazon itself—master of duplicating and undercutting whatever is selling on its own platform—won’t soon follow suit, well, think again. Books is where the online retailer got its start, and they continue to dominate in both books and audiobooks. Perhaps they’ll even lead the way.
The Human Algorithm
The algorithm has long been in place. It’s called the market. And the apparatus of publishers, editors, and agents was nothing more than the human embodiment of the algorithm, as these professionals would attempt to shape the manuscripts delivered by hopeful writers into something more palatable to said market’s maw.
My own very limited experience with the publishing world confirms this. When I started writing these books, I did not think of them as multi-book series. My agent, a man with decades of experience in the biz whose opinion I value, suggested I think of them as such. He also suggested that I end my earlier, rather gloomy and tragic versions if not with an outright HEA (“Happily Ever After”), then at least with a more positive upswing. Because this is what sells. I trusted his knowledge and experience, made changes that both he and I could live with, and I believe the books are better as a result.
Then, as we were shopping the manuscript of Amba, I had a meeting with an editor at Little, Brown. At the outset of the interview, the editor asked me whether I wanted to make this more of a brooding, literary work that might find a niche audience? (Where it was, albeit roughly.) Or do I want to write a bestseller? I opted for the latter, largely because I know that publishers want bestsellers, not niche or midlist. And because I would like to sell more books rather than fewer, certainly. I incorporated his suggestions into the manuscript as well, and I believe it’s the better book. (Though they subsequently passed, despite the tailoring.) These iterations, however, took me months. Years in the case of Amba. With the aid of AI, I’m sure that could’ve been shortened immensely. And a true AI would not have to go through any of these iterations, it would simply spit out something directly tailored to the market.
Certainty vs. Variety
The joke in the Simpsons is that the monkey, typing randomly, was only off by a few letters. Of course, the AI wouldn’t make such rudimentary spelling mistakes. (And presumably wouldn’t be trying to retype Dickens’ A Tale of Two Cities—but what if we set it in Las Vegas and Salt Lake City, in the year 2050, after the fall…) There are those who believe that there will be other, more profound “mistakes,” other tells, that will allow the discerning reader to know that whatever they are reading was written by an AI. I highly doubt it. First of all, with each new iteration, AI is improving. Exponentially.
More importantly, I don’t think that we, as readers, will care. As people, we have two competing needs: we crave certainty, familiarity, while at the same time we want variety. However the sort of variety that we tend to like—especially when it comes to media—is usually by degree, not by kind. We tend to read in certain genres. We tend to read our favorite authors. We tend to follow the suggestions the algorithm already provides, as it learns our tastes—if you liked this, you might like that. Sure, I’ll check it out.
When the second manuscript I sent my agent was a fantasy and not an action-adventure like Amba—we hadn’t yet placed Amba, and I was just writing whatever interested me—he made no secret of his dismay. It’s very hard to make a brand for yourself across genres. I like to read across genres, but most people don’t, and the real consumers—the rabid readers—are often very devoted to a certain subgenre (or two, or three).
The game—on Amazon at least—is the ability to make yourself a brand and to generate the “long tail,” which banks on the dual propensity of consumers to stick to brands they like and to binge content. That’s why there are so many series on Amazon, and why I’m writing a series: you hook readers in with an interesting and less expensive, or even free, first book—your “loss leader”—and then a percentage of those initial readers will continue on to the next books in the series. Imagine being able to spit out all of the Conan books in an afternoon or two. Or if Harry Potter ran to one hundred volumes. AI-supported writers may have to pace their offerings so as not to swamp their readers—unless they discover, as Netflix and others have, that it’s more profitable to just upload everything at once, and let the consumer binge as they are able.
As a writer well aware that each day that passes spent, say, writing about impending AI apocalypse is another day that I’m not writing my next book, the thought that AI bots could soon be swamping the market is queasy-making. I finished a draft of the sequel to Mirrenwood, only to realize that I was trying to cram material enough for two books (well, a book and a half) into one, and so I would need to restructure, and rewrite. A typical setback, but costly in terms of days and momentum and personal frustration. And I’m not a professional, I don’t have an editor demanding that I finish, nor do I rely on Amazon sales for my livelihood, the way many self- or indie-pubbed authors do. It’s hard to imagine the coming competition from a pro’s perspective.
The Best of Times? Or the Worst of Times?
I’ve seen two main reactions from writers, and I wonder if this doesn’t sort generationally. There are those who view novel-writing AI as the worst of times. How can we humans compete, once the technology is good enough to spit out novels tailored to niche markets? Or, once it becomes widely and cheaply available enough for consumers to generate their own novels by plugging in their own specific criteria? Perhaps we will learn that we truly do want to star in our own stories.
Then there are those—perhaps skewing younger and more tech-savvy—who consider the coming AI revolution to herald the arrival of the best of times. This will be a tool that we as writers can use to generate new ideas with the push of a few buttons, rather than hoarding them as precious commodities in cramped little notebooks. Or we can use it to spit out the content for us, that we as writers can then shape and craft: “elevating” us to the role of editor and publisher. It might actually be nice to not have to chain ourselves to our proverbial typewriters, to not have to sit in front of yet another screen for hours upon end. A writer could produce 100 books a year, and that might just become the new standard.
I’ve also seen the suggestion that publishers will reassert themselves, as their imprimatur will signal to readers (well, at least to Barnes & Noble shoppers) that this novel was actually written by a human, and (therefore?) is actually good. That ship has long since sailed, however. I don’t believe enough readers actually trust the “Big Five” publishers to have the sort of discriminating taste that they once had (and I mean that in both senses of the word “discriminating”). We all know that the “Big Five” publishers are now owned by even larger corporations and are driven entirely by sales and profit, not excellence or merit, and certainly not by “art.” Also, the technology will be too tempting for them not to use—they’ll be able to get rid of readers and junior editors, when they can have an AI assessing incoming manuscripts. They will not turn away authors who are using AI, so long as those authors are selling well.
I’ve seen some express the opinion that what will matter, in this brave new world, is the connection that exists between the writer and his or her readers. Which amounts to clinging to a belief in a celebrity-fan dynamic—a parasocial relationship, which with social media is quickly becoming one of the most important relationship dynamics in our culture. (Those who have this opinion often either have or are working toward building such a parasocial relationship, or celebrity/influencer status.)
But if AI can become adept at portraying believable characters—which I would argue is one of the defining hallmarks of a successful novel—then what is to prevent an AI from generating the character of the “author”? What is a celebrity but a character that we identify with? That we root for? That we “like”? AI is producing successful digital imagery, video, voice-over—who wouldn’t fall for an intimate weekly fireside chat with their favorite author where said author talks about their life and loves and hardships and formative influences, able to discourse not only on the books they have written but on any subject under the sun, easy to engage with and willing to answer intimate questions from his/her/their/its fans? Would it be such a stretch for an AI to portray this “author”? I’m sure we’ll get there. I mean, if there is money to be made from it.
The day will come when I will read an AI-generated novel or two, just to see what it’s like. But I imagine I will tend to look for books marked with a label like “AI-free content” or “No AI was used in the generation of this content.” Or just the letters “AI” in one of those big red circles with a slash through it. But, of course, if that starts to sell, then the AI and less scrupulous publishers will adopt this branding as a strategy as well, just as food brands use “all natural” or even “organic.” Most consumers don’t care whether their car was made by robots or humans, or whether the chicken they’re eating was raised free-range and humanely or lived its life in a cage, fattened by hormones.
Similarly, I didn’t pay much attention when most manufacturing jobs were replaced by robots, causing the destruction of vast swaths of this country, opening up hundreds of previously thriving communities built around factories to the perils of unemployment, the destabilization and despair that result—a much worse catastrophe, I would suggest, than the coming revolution in knowledge work. (I was still a kid, and then a callow young adult in the 80s and 90s when much of that was happening, but I wasn’t oblivious to it—it just didn’t affect me.) We knowledge workers—the journalists and writers and editors and artists and filmmakers and tech workers—whose livelihoods are being threatened by the next wave of robotization will make a much more public stink about it in the media. In the remaining time that we have access to it, at least.
For now, all I can do is keep plugging away, because it’s an activity I enjoy. I can try to write smarter, better, and faster. With a wary look over my shoulder all the while.