I was recently invited to attend a webinar sponsored by Publishers Weekly on the topic of artificial intelligence in the publishing industry. If this had come a year ago, I likely would have passed, but lately I have found out that this concept does not necessarily bring on the night terrors. AI is a colossal topic, but by breaking it down to study one facet at a time, it can be comprehensible to the tech-savvy reader.
The webinar was Artificial Intelligence: Revolution and Opportunity in Trade Publishing, presented on the afternoon of Sept. 27, 2023. The moderators were two soft-spoken gentlemen who are also very deep thinkers on automation issues. Peter Brantley is director of online strategy at the University of California–Davis library. Thad McIlroy has had a full career as a consultant in publishing technology, and he is the author of more than a dozen books and hundreds of articles and the creator of The Future of Publishing blog.
In a brief introduction, McIlroy said that in devising this event, he was more interested in hearing from people in the trenches than ivory tower theorists, and that practical slant was seen throughout the afternoon.
Opening Remarks and First Panel
Markus Dohle, former CEO of Penguin Random House, had the first word. He welcomes this new technology and said that this is the best time ever to be in publishing. He called AI the next inflection point in our digital transformation, noting that the first inflection point was the creation of Amazon 3 decades ago. The second inflection point was the Kindle, which brought about the mass market for ebooks. Digital technology has made us more efficient at publishing, he said. Copyright protection is the lifeblood of this industry, and Dohle thinks we will find guidelines that are fair to all concerned.
In the first panel, Barbara Kline Pope of Johns Hopkins University Press said she encourages her staffers to learn AI, play with it, and put it to work. She is particularly interested in them applying it to mundane tasks. When the staffers asked her for guidelines, Pope put the question to ChatGPT and got a list that did the trick. In a later panel, her Johns Hopkins colleague Diem Bloom pointed out that AI was particularly useful in indexing and abstracting.
Author Michael Bhaskar is concerned about the tsunami of data coming in from AI. He wants to know how to make a business model that protects intellectual property in this new universe. He asked people to consider what publishing even means when you can create a book with the push of a button.
AI and Book Production
The next panel featured Ken Brooks, Bill Kasdorf, and Diem Bloom. Brooks, a consultant in automation processes, got his start in AI with ChatGPT and still considers it his main tool. He said that AI will bring down the cost of production. Kasdorf pointed out that AI is already part of everyday automation tools such as Word and PowerPoint. One area where AI is lacking is the creation of alt text in Word files, which tends to come across as laughably stilted. He noted that AI narration is at a point at which it can be effective for publishers who want to create audiobooks out of their backlists. This is good news for those of us who remember text-to-voice productions from a decade ago that made everyone sound like robots.
Getting the Lawyers Involved
Scott Sholder is an intellectual property attorney. Since he signed up to be in this webinar, he was added to the Authors Guild team in its suit against OpenAI. Because it is pending litigation, he had to be circumspect in talking about that particular case, but I found a recent quote from him saying, “The plaintiffs don’t object to the development of generative AI, but the defendants had no right to develop their AI technologies with unpermitted use of the authors’ copyrighted works.” (Disclaimer: I am a paid-up member of the Authors Guild.) The suit is about material pulled from the web and put into a giant database to train AI programs how to write.
Sholder said copyright protectability depends on how much of a work was the result of human input and how much the result of machine output. Currently, art created entirely by an AI machine cannot be copyrighted. Is harvesting material from the web within fair use guidelines? To start, it depends on whether it is fiction or nonfiction. Sholder said that European courts are more aggressive in these matters.
Ethan Mollick is a Wharton School professor and the author of a very popular blog, One Useful Thing. With boyish enthusiasm, this fast-talking expert crammed a half hour’s presentation into 15 minutes. He said educators have every right to be concerned about students using AI as a crutch, since it is mostly undetectable. His well-illustrated talk contained only one photograph that is actually real, and he said nobody can guess which one it is.
Today’s AI is leaps and bounds ahead of what has been developed before, so be prepared for big changes. The most eye-opening report here was that AI programs were given standardized tests such as the SAT and the LSAT and consistently outperformed humans.
A Talk With Thad McIlroy and Peter Brantley
The day after the webinar, I was given a chance to speak directly with the two moderators. With some trepidation, I wrote a half-dozen questions for these experts, and what followed was 45 minutes of dazzling and informative conversation. Expanding on the idea of AI narrators reading audiobooks, I asked them how feasible it would be to create the voices of deceased actors. They told me that such technology is already here. A London-based company, DeepZen, has secured rights from the family of the late actor Edward Herrmann to use his voice in audiobooks. Brantley suggested that one day we will see programs that can read text on demand in the voices of actors that we choose.
I asked to hear some examples of the dangers of AI, and McIlroy told me that there was an article about it in The New York Times that very day. The ideas include dystopian views of a world in which smart computers enslave humanity and concerns that this technology will be used by terrorists. To that end, there is a call for better regulations to prevent corporate misuse of AI. Given the record of today’s Congress solving even simple problems, I’ll remain pessimistic for now.
I asked if AI can be used to answer questions beyond our current knowledge, such as solving the mysteries of dark matter or dark energy, but Brantley thinks we are not at that point yet. He suggested that I look into DeepFold, which is an AI-based program that analyses the structure of proteins in ways that leave other methods in the dust. Early in the webinar, they said that AI is still in the toddler stage. Perhaps when it is a teenager, it can solve dark matter.
I mentioned to them that as someone who grew up reading science fiction, none of the authors in my youth predicted a future in which people carry glass-and-metal objects in their pockets that amuse them and help them communicate. Brantley suggested I check out a Ray Bradbury story called “The Veldt,” which concerns a family living in a technology-infused house where the children are too involved in the gadgets.
Now It’s Getting Personal
A recent article in The Atlantic (subscription required) links to a site on which you can check to see titles from the Books3 dataset that have been harvested for the purpose of training AI programs how to write. As the author of three library-related books, I thought it would be a long shot, but I was wrong. My most recent book was there. My wife asked me if I felt used or complimented. I admitted I felt a little of both.
In 1990, when I began my career in earnest as an academic automation librarian, I was in awe that this was the time of the greatest revolution in information since Gutenberg’s printing press. Now the experts are saying that the rise of generative AI is a revolution on that order. I can’t wait to see what happens next.