At Publishers Weekly's AI web conference, Artificial Intelligence: Revolution and Opportunity in Trade Publishing, held on September 27, book publishing and technology experts alike took a long look at the fast-developing AI technologies causing a stir in the creative industries. And over an afternoon-long discussion, hosted and moderated by Thad McIlroy, digital publishing analyst and PW contributor, and Peter Brantley, director of online strategy for the University of California Davis Library, a range of speakers and suggested that despite serious questions and potential challenges, AI could augur positively for the book business.

In his opening keynote, former Penguin Random House CEO Markus Dohle kicked off the webinar by calling AI another digital "inflection point" for publishers, and sought to assure industry members that AI will not be the death of publishing. Much like the introduction of e-commerce and e-books did not only not kill publishing, Dohle anticipates that AI might, in fact, help to usher in a new and prosperous era for the business. “For me, AI is the next inflection point in the digital transformation of our industry,” Dohle said, “and the huge opportunities it offers outweigh by far the potential challenges and threats we need to meet.”

Noting that book publishing has adapted to new technology developments better than most media segments, Dohle added that the printed book remains the cornerstone of the global market, augmented by such digital innovations as downloadable audio, which has grown into a multibillion dollar business.

“Ultimately, the industry has significantly benefited from digital developments and tools—actually information technology in general,” Dohle said, pointing out that publishing has become more effective and efficient because of technology. AI, Dohle said, has the potential to improve what he sees as the two main services publishers provide: creativity, whereby publishers discover and nurture writing talent, perfect their stories, and package them before they launch them into the world; and reach, through which publishers develop a vision for the audience of a book and try to maximize readership and, with that, the selling of a story.

“It’s already clear that both machine learning and generative AI will further transform, enhance, and optimize the reach services and back-office functions of publishing,” he said. Dohle has high hopes that AI will be able to create sophisticated recommendation engines for readers. “Finding the next best read is still the most important question our readers have—and AI will help to answer that question in a much better and more personalized way,” he predicted.

Beyond that, though, Dohle argued that AI might have a fairly significant effect on the creative part of publishing, as well as on other content industries. According to Dohle, generative AI—for the first time—is able to create sophisticated design work, illustrations, and longform narratives on its own. In some cases, he said, these narratives might be difficult or even impossible to distinguish from human creation. “That is certainly something we need to carefully navigate and regulate in order to protect our authors and their works—and human creativity in general,” Dohle said. “Copyright protection, and with that author compensation, is the lifeblood of our industry. If we lose that, we lose everything!

Dohle said that he is confident that publishing will establish the necessary legal framework and governance standards to “A., support and fairly compensate human creativity for longform storytelling in the future, and B., benefit from this next iteration in the development of information technology that will result in significant improvements and efficiencies in the services we provide as publishers.”

Book Publishing Adjusts

In the panel immediately following Dohle's talk, on "AI and the Business of Publishing," the three speakers agreed that AI will likely make more of an impact on publishing than e-books did, with author and publisher Michael Bhaskar noting that AI, in the broadest sense, is challenging the existing foundations of such things as our culture and laws. He called the development of AI “an earthquake moment.”

Cathy Weldon, an AI and machine learning associate at Penguin Random House, said that AI is impacting all areas of publishing, including on the creative side of the industry. She added, however, that AI is “not that new,” observing that she has been using AI to help make print forecasting decisions and create book recommendation programs at her job for two years.

Barbara Kline Pope, executive director of Johns Hopkins University Press, said she hasn’t been this excited about a new technology since the creation of the internet. Kline Pope said that she sees AI’s place in publishing as helping people to be more creative, not as a way to replace employees. She has encouraged her staff to experiment with AI, noting that, somewhat to her surprise, employees who experimented with the technology did not shy away from its use, but rather requested guidelines on how to best use it. Kline Pope said she has drawn up a set of those guidelines around such things as how to protect the press's content.

Bhaskar agreed that having staff experiment with AI is a good way to help them overcome their negative view of the technology. He sees determining the proper uses for AI and developing strategies to reimagine business models as publishing management's main job in response to the new technology.

Weldon firmly endorsed the need to develop guidelines, and said that, as the hype around AI has spread, she has had more interactions with different groups within PRH who are curious about the technology and who have questions. “It feels good to be more connected” to other parts of the company, she said.

Weldon was the first panelist at the event to raise concerns over the use of large language models such as Chat GPT. She warned against using them without human oversight, observing that most LLMs “have factual inaccuracies.” Kline Pope saw a more insidious problem, noting that many LLMs come with biases that can quickly spread, to the detriment of marginalized groups.

Bhaskar pointed to the fundamental challenge facing publishing about AI: everyone knows it is coming, but there has yet to be any agreement on how to regulate it, or monetize it fairly. “It is an IP challenge without precedent,” he said.

AI and Copyright

A pair of late afternoon panels touched on the anxiety AI is causing among creators—specifically the copyright concerns tangled up with AI companies training their large language models on troves of unauthorized, and in some cases pirated, content scraped from the Internet, and the potential for those AI services to then put the creators they were trained on out of work. Numerous class action lawsuits have already been filed by creators claiming copyright infringement, including four separate actions filed by authors.

Fair use is a very, very grey area. People have strong opinions on both sides. My clients have as you know have taken a stance. There will be strong opposing viewpoints to that position, I'm sure. And a court will have to decide.

One of the attorneys representing one of those author groups, Scott Sholder, litigation co-chair at law firm Cowan DeBaets Abrahams & Sheppard, spoke with Brantley about the intricacies of copyright law, fair use, and fair compensation for creators—and how AI represents a unique challenge. While much of the dialogue revolved around the potential complexities of using AI as a tool for creation, eventually the conversation came around to litigation and fair use, and whether AI’s ability to, as Brantley put it, take in all these “grains of sand” and turn out a “sand castle” crosses a legal line somewhere.

Sholder, citing his current case, demurred. “It’s a little bit hard for me to go into much detail on this because [fair use] is going to be a significant issue in many if not all of the cases that are that are pending, including my own,” he acknowledged. “Fair use is a very, very grey area. People have strong opinions on both sides. My clients, as you know, have taken a stance. There will be strong opposing viewpoints to that position, I'm sure. And a court will have to decide.”

Immediately following Sholder came perhaps the day’s most insightful, searching discussion, with Brantley talking with two authors who have both experimented with AI: Sean Michaels, author of Do You Remember Being Born, and bestselling author and screenwriter Gregg Hurwitz. Both authors spoke of approaching AI technology more with artistic curiosity than with fear.

“I believe very strongly in the power of creative people, of artists, painters, writers to use whatever tools are at their disposal, just as I believe in their power to draw inspiration from whatever forces are in their life to make inventive, creative, beautiful work,” Michaels explained at one point. “It's hard for me to imagine a world 10 or 20 years from now where another generation of artists isn't making incredible work that is truly human-driven but that draws on these tools in interesting ways.”

Hurwitz agreed. “What people want to see, all the time, is going to be human excellence,” he emphasized. “People didn't want to watch Deep Blue play a chess match against Deep Blue. They wanted to see how Kasparov would fare. No one wants to watch an AI generated basketball game. We want to see Michael Jordan soar.... What people are interested in seeing is a version of human excellence.”

In his closing keynote, Ethan Mollick, Wharton associate professor and author of the One Useful Thing newsletter, fully endorsed the idea that the best way to get to know what AI can do is simply to use it. If you dive into something, he said, you will likely learn things no one else knows, since no one really knows all that much about AI in the first place.

At present, Mollick noted, AI is undetectable, ubiquitous, and transformative, and the only thing that is predictable is that AI will constantly change in the years ahead. But how quickly it will change, he added, is impossible to predict. Among his principles in dealing with AI: 1. invite AI to everything, 2. be the “human-in-the-loop,” and 3. tell it who it is (and treat it like a person).

The webinar will remain online through October 27. Tickets can be purchased here. And PW has set a second half-day webinar for February 21, 2024: AI Tools and Solutions for Book Publishers: Harnessing the Power of Artificial Intelligence. Look for more details soon.

Return to the main feature.