AI and Book Publishing: The Use Cases

Keith Riegert, CEO of Ulysses Press and Perfect Bound, presented at the Publishers Weekly U.S. Book Show in May, 2024, offering the most comprehensive overview I’ve seen about AI use cases within publishing companies. Perfect Bound is a sponsor of this report; I stand by that statement. He offers “20 practical ways you, as a publishing professional, can start using AI right now.”

His presentation, Getting Started with AI, can be viewed and downloaded from the Perfect Bound website.

I’m borrowing this section title from Ethan Mollick’s newsletter deliberately—there’s no need to try to improve on it. Mollick is a professor at the Wharton School of the University of Pennsylvania, who studies entrepreneurship & innovation. His newsletter, which I recommend frequently, is calm, refreshing, and uniquely insightful.

Among the things that qualify Mollick as a commentator are that he has no skin in the game. He doesn’t need to sell AI, nor to trash it. He has merely committed to exploring AI in its many impacts, mostly upon education, culture, writing, and publishing. And he’s a fine, clear writer.

If you browse through Mollick’s newsletter archives you’ll see that he didn’t begin to focus on AI until December 2022. It wasn’t his beat—like most of us, AI dropped in on his day job, and he couldn’t take his eyes off it.

In this post his insights falls closest to our interest as publishing professionals. “Might AI,” he asks, “change the way we interact with books?”

To answer the question, Mollick notes, “we need both an AI with a memory large enough to hold a book, and an author who knows their own book well enough to judge the AI’s results.” Mollick tests one of his several titles (he doesn’t specify which one, but from the chats it’s clearly The Unicorn’s Shadow: Combating the Dangerous Myths that Hold Back Startups, Founders, and Investors, a book favorably reviewed on Amazon, though not a current bestseller).

Mollick considers different aspects of AI’s potential value to an author, publisher or reader, including “AI as reader and editor,” and “a practical use: help for instructors.” He asks an LLM—large language model—not ChatGPT—to summarize the book. It succeeds to Mollick’s satisfaction.

Then a tougher challenge: “Give me examples of metaphors in the book.” Metaphor, he points out, “is challenging for even human readers, as it involves finding a use of figurative language without any clear markers (unlike a simile, there are no “likes” or “as”).” The results, he records, “are impressive, though there are minor errors.”

The LLM is less successful as an editor: its failings in this department, Mollick notes, illustrate “something that has become clear about the current state of AI: if you are a very good writer or editor, you are better than current AI…”

Nonetheless, “AIs have, or at least have the appearance of having, an understanding of the context and meaning of a piece of text.” As a result, Mollick believes that “how we relate to books is likely to change as a result of AI.”

I think so too.

Software for the automated typesetting of books dates back to at least the 1970s. In the mid-1980s I supervised a software project called PageOne, based on Donald Knuth’s TeX, which could typeset a book in minutes. SGML appeared around the same time, based on a document standard introduced in 1969. It was largely succeeded by XML, introduced in 1996. These robust markup languages create solid structures for automation.

Desktop publishing ushered in another round of automation for QuarkXPress and Adobe InDesign, as well as Adobe Illustrator and Adobe Photoshop. Publishing workflows can be managed with various programs and systems.

An organization to watch is the Coko Foundation. They offer a suite of open source production and publishing management tools, including Kotahi, a scholarly publishing platform, and Ketty for book production, which includes an AI Assistant. The Kotahi AI PDF Designer, “transforms PDF design into a straightforward, interactive process.”

There are some early initiatives to bring AI into InDesign workflows. In April 2024 Adobe announced a Text to Image feature. Third parties may be getting ahead of Adobe here: the innovative prepress and production vendors in India, such as Hurix Digital and Integra, are showing more initiative than Adobe in harnessing AI for production.

The ‘low-hanging fruit’ is obvious. Ask Chat AI to help with a product description or a press release. Ask it to suggest some keywords. This it can do, without breaking a sweat. But most publishing professionals can do the same thing, with only a little moisture on the brow.

Keith Riegert’s use cases, linked above, include suggestions for brainstorming titles, drafting a digital marketing report, and creating a digital marketing campaign tracker in Google Sheets.

Shimmr software, described above, hints at the shape of automated marketing to come.

Metadata is core to book discoverability. You’ve heard that enough times to be nauseated by the admonition. It’s off-putting mainly because “metadata” remains elusive to most non-techies. If you say, “it’s just the basic info about the book, the title, description, price, subject categories, that kind of thing,” people exhale. That they’re comfortable with. But that’s about all.

I regret to remind you that there’s actually far more to metadata than just a few details about the book. There’s so much more. Much more than I can encompass in this little book. I’ve co-authored a whole book on the topic. Ingram publishes Metadata Essentials, an excellent short volume. I’ll say it here, and not for the last time: authors and publishers pay short-shrift to their metadata at their peril.

AI can help with metadata generation. For example, self-publishing vendor PublishDrive, offers an “AI-Powered Book Metadata Generator” which offers AI recommendations for the book title, blurb, Amazon categories, BISAC categories, and keywords.

Insight, from Veristage, described above, can generate descriptions, keywords, BISAC categories, and define target audiences.

With his typical deep wisdom, Graham Bell, the organization’s director, notes that “one reaction to (the controversies surrounding the technology) is to forswear use of AI or to avoid trading in AI-created products. A more realistic option is simply to be transparent with trading partners and readers when AI has been used. And as some resellers limit or ban AI-based content from their platforms, it is important for reputable publishers to highlight those products that do use generative AI techniques to create content.”

Bell goes on to outline ways that publishers can specify in metadata:

• AI contributors

• AI-based voices in audiobooks

… as well as a method to indicate in the metadata for digital products that the publisher explicitly opt outs of text and data mining (TDM) for uses other than research. There’s also a way to specify a separate license covering commercial or non-research TDM.

As is often that case, what is specified in ONIX may not be uncovered down the food chain, but at least a best effort has been made.