The latest generation of AI is a game changer. Not incremental change—something gentle, something gradual: this AI changes everything, fast. Scary fast.
I believe that every function in trade book publishing today can be automated with the help of generative AI. And, if this is true, then the trade book publishing industry as we know it will soon be obsolete. We will need to move on.
There are two quick provisos, however. The first is straightforward: this is not just about ChatGPT—or other GPTs (generative pretrained transformers) and LLMs (large language models). A range of associated technologies and processes can and will be brought into play that augment the functionality of generative AI. But generative AI is the key ingredient. Without it, what I’m describing is impossible.
The second proviso is of a different flavor. When you make absolutist claims about a technology, people will invariably try to defeat you with another absolute. If you claim that one day all cars will be self-driving, someone will point out that this won’t apply to Formula One race cars. Point taken.
This isn’t about Formula One publishing. I’m going to be talking about “good enough”—about what people will accept, what they’ll buy, and what they’ll actually read. I’m not going to claim that Formula One publishers won’t be able to do a better job than AI on many of the processes described below. But I’ll challenge you to consider exactly where the human touch brings sufficient added value to justify the overhead in time and costs.
Does any of this mean that there will be no future for great novels and fine nonfiction studies? Of course it doesn’t. That’s not my point.
Do I doubt that there will still be fantastic cover designs from talented designers? Of course there will be. We’ll still stumble on new books on bookstore shelves and, humbled by the grandeur of their cover designs, declare that there’s no way they could have been designed with AI. And sometimes we’ll be right.
My reference point is what we lived through with desktop publishing. Many of the people working in publishing today entered the industry after 1985, when the combo of the Macintosh, the laser printer, and desktop publishing software were first deployed. The initial results were not pretty. The type was rough around the edges, and the letterspacing was crude.
In 1988, when the Macintosh was hooked up to a Linotype machine, the improvement in quality was dramatic. But most traditionalists still argued that the “color” of the type remained poor, and that this lack of quality would be perceived by readers and rejected. The specialists—the designers and their publishing clients—had built careers in part around paying attention to the nuances of specific typefaces, and to kerning, line spacing, and the design of the printed page.
It’s not that this became unimportant, or that there was no appreciable difference; it’s that the new technologies could produce “good enough” quality—a new concept for measuring publishing output: namely, good enough for the vast majority of readers to be perfectly happy with what they were seeing and reading in the books that they purchased.
Acquisitions and editorial
Professional copyediting is the kerning of 2023. The tech is not quite here today. I don’t think that GPT-4 can yet handle copyediting to the standard that book publishers require. But that ability is going to be here sooner, not later. While professionally copyedited books may still be “better” to a refined editor’s eye, you won’t be able to sell more books with the professional human touch. They will already be good enough.
What about developmental editing? You might not let a GPT make the final editorial decisions, but you’d be foolish not to ask it for suggestions.
And ChatGPT will become the patron saint of the slush pile. Its abilities to evaluate grammar and logical expression allow it to make a once-over assessment of whether a book is (reasonably) well written. It might not spot the gems, but it will know how to separate the wheat from the chaff. Ah, you will say, recalling one of those manuscripts that were rejected by 100 publishers but went on to become an unexpected bestseller—surely a GPT might miss those, too. Yet so did 100 purportedly well-trained publishing professionals.
Most book production, print and digital, is already fully automated or semiautomated, mostly rules based. While still evolving and improving, automated page production has been employed for decades—Donald Knuth’s “programmable” typesetting system, TeX, first appeared in 1978. Much of the output of Adobe’s InDesign is automated through macros and scripts. Single-source publishing systems are securely in place.
AI will fill in some of the few missing pieces, but production inefficiencies exist largely because publishing is still, by choice, a high-touch process, based on the largely unshakable belief of many in our industry that the value of human intervention outweighs the lost productivity from automation. This is unquestionably true for multiple categories of publishing. It’s unquestionably false for others.
Distribution and advertising
For the publishing industry, online distribution and advertising have separated writers from readers. Self-published authors have proven that the closer one gets to their audience, the more fans they will get and the more books they will sell. While online resellers aggregate audiences into big broad buckets, AI disambiguates them, enabling writers and readers to forge direct connections.
Amazon has become an overpriced rentier that publishers can ill afford. It can still be a door opener for new authors, but for established publishers it charges too much for what it delivers.
Amazon’s dominant position in bookselling is not going to change overnight, nor even in the morning. But part of the publishing transformation that AI will engender will be a series of energetic attempts to disrupt Amazon’s position in the distribution ecosystem. As media continues to morph, AI seeds new delivery channels. Amazon will try to dominate each new channel via acquisitions, as it did so brilliantly when it bought Audible in 2008 for $300 million. But Amazon is a lesser player in the video and gaming spaces, and, as yet, in the new entertainment channels that AI is germinating. This is shaping up as a classic example of Clayton Christensen’s The Innovator’s Dilemma: When New Technologies Cause Great Firms to Fail.
But I see a bright future for bookstores. It can be chilly in AI’s uncanny valley, and bookstores will remain singular sources for camaraderie and the human touch.
Marketing and discovery
Marketing will evolve and prove to be the most powerful facet of AI’s foray into book publishing. This is where the stakes are highest. One thing that could hold it back is a lack of widely available real-time sales information. Subscribers to BookScan will have an edge. It will be fascinating to see what Ingram does to enhance its Marketing Insights product, and what Firebrand does with its Eloquence on Alert.
GPTs can do a great job with competitive analysis and can paint a compelling real-time picture of what’s happening in the market to the books that are siphoning off sales, and to opportunities missed. It’s also going to deliver on the till-now thwarted promise of efficient discovery. Writers will pinpoint their ideal audience and readers will pinpoint their perfect next read.
That’s the tip of the iceberg. Write down the myriad other sales and marketing functions on the back of a napkin and then ask yourself, what might a GPT do to help?
Think about the transformation of educational publishing over the last two decades. The pursuit of education, training, and learning has not declined in that period; it has only grown larger—every day more and more people are becoming trained and discovering new skills to further their careers, or just out of personal interest. But so much of that is taking place online, on YouTube, Udemy, Coursera, and any other of the hundreds of high-quality and low-cost online resources. (And these too are being disrupted, and mostly enhanced, by GPT technology.)
These developments have had an enormous impact on the industry we call educational publishing, which has shrunk in the past two decades by double digits. Surely a new round of technological disruption could shrink the existing trade publishing industry.
Consider also the growth of self-publishing as a whole. Because the sales data mostly hides under the table, we tend still to understate its impact. But self-publishing has siphoned some billions of dollars per year away from traditional publishing companies. Read ’em and weep.
The entertainment industries
Don’t take your eyes off the entertainment industries at publishing’s periphery, and the impact that this same technology has on them. Movies, games, music, and social media more broadly—all of these industries are being impacted on a scale similar to the publishing business. And these activities are more important to most adults than reading. Time-use studies show adults spend less than 5% of their leisure time reading, and an ever-increasing percentage of their time gaming and watching videos online.
Enhanced e-books never took off, but sales of books enhanced into audiobooks have exploded beyond anyone’s most optimistic projections. Movies have been books’ love children for more than a century, with some revenue sharing, though that marriage was never fully consummated. Will it be now? Movies and video games have joined near the hip, with book publishing at the periphery. Will that change?
AI is going to enable books to morph into additional revenue-producing mediums, in ways we’ve never seen before.
The perils of AI
I want to put this bluntly. All of this talk about the problems and perils of AI is just a distraction. It’s a distraction from understanding the opportunity that AI brings to book publishing. This is not in the least to undermine the discussion, nor to in any way to suggest that the concerns are invalid or somehow unimportant. But you must fill two deep buckets when you think about AI, promise and peril, and keep a deep moat between them.
If you try to evaluate the opportunity through the lens of the coming AI apocalypse, you’ll have a hard time seeing the green light ahead. You could argue that we as an industry must assess the coming dangers in order to make the hard choices about how AI should be deployed in the publishing industry. Yes, we should make those assessments, and we can make pronouncements. But they will be unenforceable (except, occasionally—and eventually—by law). Entrepreneurs will reach into our industry however they choose, picking off the low-hanging fruit, regardless of the rules that we who are in charge seek to impose. (Did self-published authors listen attentively when we said, “No, you really can’t do that”?)
I’ll take a libertarian tack: you can only understand the perils surrounding a new technology after you fully appreciate the opportunities that it affords.
I’ve argued (briefly, superficially) that every function in trade book publishing today can be effectively automated with the help of generative AI. If this is true, trade book publishing as we know it will become obsolete. It’s not that books won’t be written and published and read; it’s only that the mechanisms by which they are developed and sold will largely be machine driven, and so the publishing industry will have a very different form. New opportunities will appear behind the fast-closing doors.
“The biggest change to hit publishing since the invention of the printing press” is a hoary old cliché that tries to evoke the scale of a publishing innovation. But is this, finally, the big one? At last fall’s PageBreak conference, Tim O’Reilly, discussing GPTs, said we’re at a point that’s “very similar to how I felt when we discovered the World Wide Web in 1992,” and followed that bold statement with “this is as transformative as VisiCalc, the PC, and the web browser.”
We know how the world was transformed after the invention of the printing press. What will publishing look like after generative AI has been fully assimilated into our current workflows? What will reading look like when books are diminished by GPTs’ I-can-answer-that-question-right-now capabilities? When it can spin a tale that delights?
I don’t shudder to think about this. I couldn’t be more excited.
Thad McIlroy is an electronic publishing analyst and author, based on the West Coast and at his website, The Future of Publishing. He is a founding partner of Publishing Technology Partners.