In this podcast episode, we share some snippets from last year’s Tech Forum sessions on AI in the book industry.

(Scroll down for a transcript of the conversation.)

Want to make sure you never miss an episode of the podcast? You can subscribe for free on Spotify, iTunes, Pocket Casts, TuneIn, or SoundCloud.

Transcript

Adaobi Nnaobi: Welcome to the BookNet Canada podcast. I’m Adaobi Nnaobi, the Marketing & Research Associate and the host of this month’s episode. Last year at Tech Forum, our educational conference on data, technology, and collaboration in the Canadian book industry, we had a series of three sessions on AI in the book industry. In this podcast episode, we’re going to share snippets from the sessions to get a feel for how AI can be used in publishing and bookselling and how to take an ethical approach to using AI.

In the first AI session, Applying AI to publishing: A balanced and ethical approach, George Walkley, a publishing veteran and leading authority on AI and its applications, covered the opportunities and challenges that generative AI introduces, offering actionable tips for staying ahead in a dynamic landscape while emphasizing ethical considerations. Here’s George Walkley on publishing use cases.

George Walkley: Where do I see AI being used today? Across all of the businesses that I've worked with, across all of the types of publishers that I've worked with, these are the top ten things that I see AI being used for. Ideation and coming up with initial ideas. Most of us are quite fearful, I know I'm fearful, of a blank screen or blank sheet of paper. It's much easier to edit than it is to create. And so asking AI for first ideas on a topic is quite powerful. Similarly, generating summaries, typically going from long documents to short ones, very powerful. Metadata creation, taking an entire book and identifying keywords, identifying themes, identifying concepts that can be put into metadata, very powerful. Repurposing content. I was at a publisher recently where the same underlying information about the book was being presented to, respectively, booksellers, librarians, school teachers, parents, and the children who would eventually read the book. It was being repurposed multiple times, and AI is very good at that kind of quick repurposing, always with a human being checking it. Content creation is another really interesting one. The book itself, I don't see any widespread use of AI to actually write the actual book. There are just too many gaps to fill in for a probabilistic model to do that well. But on the other hand, what I see AI being used for is things like chapter comprehension quizzes, lesson plans for teachers, reading group guides for literature so you have some suggested topics to talk about with your friends. Ancillary videos to support the book. All of those are being produced through AI in large number. I spoke to an educational publisher here in the UK that I was working with. They encapsulated this shift for me by saying that, prior to AI, their ambition was to have a lesson plan for every book that they published. And now, using AI systematically, their ambition is to have a lesson plan for every chapter of every book that they publish.

Content translation is really interesting. Actually, having shown a friend who speaks very good German, the AI-translated monograph that I bought was pretty good. There are real limits, I think, to this. I think shorter monographs, I think sales material and marketing material can be translated quite well. I would be very nervous about trusting an algorithm with 80,000 words of book. But on the other hand, I see translators who are using AI to do a first pass that they then quality check. And that to me feels like a very good model for doing this. The creation of audio and video is happening a great deal. I think this is really powerful, particularly for audiobooks as an accessible format. We have a zero-to-one problem with audiobooks. It would be preferable if every audiobook was voiced by a human actor. But the reality is there are so many titles in the backlist that that won't happen. But AI has the potential to open that up and to make more accessible resources. The anti-penultimate point, AI is really helpful for helping to automate tasks. I've seen VBA macros in Excel, Python scripts for data analysis, even JavaScript extensions for Creative Suite, all being written by people giving plain language instructions and then getting code that they can test and use. Within design teams, Adobe Firefly is pretty much business as usual for the generation, not of complete covers, but certainly of stock imagery that can be used as part of those covers alongside a human designer. And finally, I'm seeing custom GPTs being used more and more in publishers. The publisher that built a custom GPT for their entire backlist to help identify backlist titles that had relevance to news stories as a publicity angle. The publisher that built a custom GPT to help with font licensing. Or the publisher that put all of their sales data into a custom GPT and now have it answer questions that are phrased in plain English rather than having to filter a complex spreadsheet.

The impact of all of these use cases is immediate and in many cases measurable. And I will say that it is particularly impactful within smaller publishers. Big publishers typically have large teams, large structures, and there's a lot of institutional inertia which they have to wade through. But very often I've seen small teams of four or five people have the productivity of a team of six or seven people by using AI alongside their work sensibly.

Adaobi: Watch the full presentation for more information on copyright and fair use, the environmental impact of AI, principles for balanced and ethical AI use, and more. The next session on AI was AI for enhanced discoverability and user experience in online bookstores. In this session, Christian Roy and Jean-Benoît Dumais delved into the practical applications of AI technology in the world of online bookstores. This session highlighted how AI innovations are being employed on leslibraires.ca to improve SEO, enhance discoverability, and optimize user experience. The session explained how the TAMIS software, which uses natural language processing techniques applied to books, is being used to achieve these results. Here is Jean-Benoît Dumais on AI and online bookstores.

Jean-Benoît Dumais: Merely having an online presence is no longer sufficient for booksellers in Canada. The digital landscape presents both challenges and opportunities. According to the More Canada study that was published in 2018, changes in how Canadians discover, purchase, or borrow books are eroding the presence of Canadian expression in our cultural landscape. This concerning trend highlights the crucial role that independent bookstores can play. To counter this erosion, independent bookstores can play two complementary roles, raising awareness among readers and ensuring access to Canadian books. However, translating the unique atmosphere and curation of physical bookshops to the digital realm is not straightforward. To address this challenge, our bookstores are embracing cutting-edge technology. To make Canadian books more discoverable online, our bookstores are innovating by using AI to enrich traditional metadata. This is a task that is not performed by Amazon or other big players. As you may guess, implementing such a system requires cooperation. By fostering this partnership between bookstores and publishers, we are moving to the forefront of digital innovation in our book industry.

Adaobi: Watch the full session for Christian Roy’s presentation on how AI helps with metadata, discoverability, and user experience using AI tools like the TAMIS software and the robot reader. The last AI session was Writing, editing, and designing books with AI-enhanced tools and workflows: Lessons learnt in a rapid content-creation context. In this session, Dione Mentis from the Coko Foundation and Barbara Rühling from Book Sprints discuss how AI-enhanced tools are transforming book production workflows. The team at Book Sprints uses Ketty, an open-source, AI-integrated platform developed by the Coko Foundation, to guide writers through a collaborative, rapid content-creation process. As part of that process, writers build their own knowledge base and develop custom prompts for content creation and editing, tailored to their style guides. Here is Barbara Rühling on how Book Sprints is integrating AI into their workflow.

Barbara Rühling: Yeah. Hi, my name is Barbara. I am CEO of a little company called Book Sprints, where we help groups of experts to create nonfiction books in five days. So, Book Sprints was founded by Adam Hyde, who also founded Coko Foundation. So, we've been working with Dione for the last year and a half or so on thinking about workflows, how to integrate AI into our five-day process of collaborative book writing. And, yeah, I think we've all by now understood that it doesn't happen that we ask AI to just create a book for us with the press of one button but rather breaking it down into very little steps and tasks and see, where can AI enhance the process and help us move along.

So, now that we started exploring how to enhance our processes with the help of AI, our assumptions in the beginning were: Suddenly at every Book Sprints that we have, we have something around maybe 10 to 20 book sprints a year, everyone is going to be using AI tools. So, we thought this is going to be wild. How are we going to deal with this? We need to get ahead of the game and we need to figure out how we will handle the situation. What we found in reality is that the experts from many different fields that we've been working with have been very hesitant to include AI in their processes. So, we actually found that there has been at first very little use at all. And then when we saw it slowly coming into the process, we saw that most experts have been using those tools very mindfully. So, again, one of our assumptions has been that, "Oh, this is going to be very inefficient. People will start using, you know, ChatGPT, for example, for the first time. We'll be able to create long, long, long texts, and then somebody has to go through and edit all of this and make it work." And we really need to be able to control this. But instead we found most experts have been quite mindful about how to integrate it in small ways and, yeah, keep it under control.

And then our third assumption was that collaborative writing and the way that we do it needs very specific approaches. It's different from an individual author having that direct conversational style as we see in ChatGPT, for example. Instead, we need to find out, how can we still create cohesion and consistency when there is already a group of diverse voices with different writing styles? So, that is what we've been working on at Book Sprints for the last 15 years. How do we make this work with different people who bring in a diversity of perspectives but also obviously sometimes conflicting opinions, different writing styles? And how can we do this now that AI is involved?

Then we also thought about what is our approach to using AI. I mean, there's organizations who reject the use of AI, and I understand, of course, the reasons for it. And there's others that wholeheartedly embrace it. And we thought, well, for us, it's important that the use is ethical and transparent. So, that's very hard to define because at the moment, the legality is still being discussed a lot. It's very complex. It's also regionally different. So, to the best of knowledge and according to sort of fair use, how can we use it in a way that seems to be ethical. And transparency in the sense that how can we disclose how AI-enhanced tools have been used. We looked at a lot of different examples. How do other publishers deal with it? And some have a very minimum standard of disclosure. So, for example, Amazon now for Kindle, they say anything that has been supported with AI tools, there is no disclosure necessary. If it's really been created from the bottom up, then there needs to be a disclosure. Whereas, others are very strict and say every single prompt that you've been using should be cited, should be reference. So, we looked at sort of different ways of how can we find a text that the experts can agree to use and add to their books and make transparent how they use the tools.

Another approach for us is productivity. Obviously for us to make books in five days, productivity and efficiency is something that's been on the top of our agenda. And we didn't want this to become a playground of, like I said earlier, people just creating loads of text and then it becomes really out of hand. Everybody has to then weed through what's even useful. For us, it was important that we make it productive and efficient from the very beginning by focusing on use cases that are small and really controllable. What are small tasks that we can enhance with AI without letting it go out of hand? And to us, that also meant that what's really interesting for us is not to replace any of the subject matter experts or the experts from our team but rather how can we free up the human writers from some tedious tasks. Like, what are little things that are very repetitive that we all know that's like, "Oh, I wish I could automate this somehow," and make it more productive in that sense.

Accuracy, of course, is a topic that always comes up. We all know by now that large language models hallucinate. So, we have to always make sure that what comes out at the end is still accurate and trust in the same manner. So, in an earlier talk in this webinar series, I heard George Walkley talk about humans always in the loop, and I think this has also been our approach. So, there has to be always an interplay between the humans and the process and then a step where AI tools come in and then the next step humans have to check this again, etc.

It also meant that we wanted to focus on using existing materials that the experts of each group and their organizations already bring in. So, Dione presented the knowledge base so we can fill that knowledge base with existing written materials, blog posts, previous books, anything that has been and base the knowledge creation on that rather than on everything that's out there. And we also decided to focus on low-risk use cases, so something that makes it easier to control and to fact check rather than creating, sort of, a long list of new content that then are hard to verify.

Adaobi: Watch the full session for Dione’s demo on how AI works to enhance the publishing workflow using the Ketty Platform and lessons learned from Book Sprints' experimentation with using AI in their workflows.

I hope those three sessions helped with some questions you may have around AI in the book industry but if you’re looking for answers to other questions on AI in publishing, we’ve got answers. In our upcoming Tech Forum session, join George Walkley for a follow-up to his presentation, Applying AI to publishing: A balanced and ethical approach. George will give a brief overview of developments since that presentation, followed by your chance to ask pressing questions about AI’s impact and potential applications in the book industry. The format for this session will allow attendees to unmute and actively participate in the conversation, creating a unique opportunity for dynamic discussion and knowledge-sharing. The session will take place on February 20 at 2 p.m. Eastern Time. You can register for the session titled, AI in publishing: Your questions answered at bnctechforum.ca/upcoming-sessions.

Before I go, I’d like to acknowledge that BookNet Canada's operations are remote and our colleagues contribute their work from the traditional territories of the Mississaugas of the Credit, the Anishinaabe, the Haudenosaunee, the Wyandot, the Mi’kmaq, the Ojibwa of Fort William First Nation, the Three Fires Confederacy of First Nations (which includes the Ojibwa, the Odawa, and the Potawatomie), and the Métis, the original nations and peoples of the lands we now call Beeton, Brampton, Guelph, Halifax, Thunder Bay, Toronto, Vaughan, and Windsor. We encourage you to visit the native-land.ca website to learn more about the peoples whose land you are listening from today. Moreover, BookNet Canada endorses the Calls to Action from the Truth and Reconciliation Commission of Canada and supports an ongoing shift from gatekeeping to spacemaking in the book industry.

We'd also like to acknowledge the Government of Canada for their financial support through the Canada Book Fund. And thanks to you for listening.