I was obsessed with the concept of Books as a Service when it couldn't work. Is it now possible with GenAI?

Publishing continues to be my apple's eye in terms of interest. While relatively out of the industry - though always with a toe inside, right Andrew Savikas ? :)-, I continue reading and listening to the smart people there, always a little frustrated by how it is the best example of the fact that industries only evolve when fully threatened, but amazed and thrilled at how some "crazies" continue pushing the limites.

I had the chance to listen to Bradley Metrock 's interview with Thad McIlroy of Future of Publishing fame, Jens Klingelh?fer , co-founder of Bookwire , (a leading ebook distributor) and Joe Wikert , who led the wonderful Tools of Change conference along with Kat Meyer , and now at Revenue Path. The conversation revolved around Thad's insightful post at Publisher's Weekly from last June about how AI is finally going to be taken into account by the publishing industry as a whole, as GenAI affects the whole value chain.

There was one interesting part of the interview regarding intellectual property and how someone like Scribd, Inc. (an ebook subscription service) or other players (maybe like Bookwire) can take advantage of how important IP is now to huge LLM builders (OpenAI, Meta and the like) to bring a structured new way of licensing IP for LLMs to the table. Even publishers with interesting and specific IP (e.g. war history book publishers) could build their own "GPTs".

The discussion reminded me of one crazy idea I had around 2012-2013 around the concept of Books as a Service. As 24symbols Online Reading , my ebook subscription service company was growing at that time, I was thinking deeply about potential scale beyond pure user growth. The same way (if at a much lower scale) Amazon was able to build AWS as a lateral business that grew to become a behemoth, I was really bold on finding value on the huge treasure box of content and metadata we had because of the hundreds of thousands (and later on, millions) of books our publishers were letting us process to offer our service worldwde. What if we could build a flexible API so that third parties could take advantage of that content? What if a startup wants to build a book recommendation engine? Does it have to grab deals with the same twenty thousand publishers we reached deals with before focusing on their business layer? What if they could just use our API as a service that could provide data and machine learning services for them? Publishers would obviously get a share of the revenue, startups and other companies could build value-added services in a much faster way, and we could enhance our revenue sources. I was so bold I even gave talks and wrote academic papers about it.

Of course, I wasn't able to build it. The day to day beat me up and soon I realized we would not have the resources to create this thing. On the other side, there was clearly no appetite on the publishers' side. When my partners and I built Quantified Reading , the concept of Books as a Service was also part of the vision. Understanding how readers read, plus deep processing of both the content, the metadata and other enriching sources could be used to build an incredible platform to create new ways of learning and enjoying the written content. There we saw much more traction with great partnerships like with Hachette Book Group and Maja Thomas , and I had the chance to continue talking about it at exciting events such as 2019's Confluence conference in London (thanks to Byte The Book !) but we still saw a huge gap between expectations, machine learning capabilities and the cost required to build those services.

Now that GenAI has decreased the effort to build many of such systems, - just look at the lastest announcement by OpenAI which enables finetuning the GPT3.5 Turbo model with local content at a really low price- is it the time now? I just went to watch Christopher Nolan's Oppenheimer and while I enjoyed it, what I would really like to do now is take my copy of Bird and Sherwin's American Prometheus (one of the books the movie is based upon) and ask it questions about what I thought about while watching the movie, and not having to re-read the 600+ pages again (mainly because I opened the book the other day and the font size is just not compatible anymore with my eyesight :D). Would someone like Scribd be able to find a common ground with publishers to start building an API infrastructure so that newer services could be offered to final users? Could someone like Bookwire, with an impressive distribution engine around the world, and that has already shown interest in thinking outside the box, be up to it? Or are publishers with a knack on innovation be building topic-specific LLMs that could become useful tools for readers that want to "talk to" the books?

It's all about finding a sound business model. Would users, companies, ... be willing to pay for this? Is it the time for Books-as-a-Service, even if we know have to call it "Talk to your books via this API"?

Or maybe, just maybe...

Cover image generated with MidJourney.

Simpson's image: https://knowyourmeme.com/memes/old-man-yells-at-cloud

#ai #genAI #booksAsAService