Two Philosophers Bring Expert-Based AI to Your Reading Experience
John Kaag (University of Massachusetts, Lowell) and Clancy Martin (University of Missouri, Kansas City) have teamed up with businessman and philosophy enthusiast John Dubuque to create a new business that brings together great books, expert commentary, and artificial intelligence.
[“What Have We Become?” vol. 3 by Nicholas Galanin]
Each of the experts, selected by Kaag and Martin, provide commentary on and are interviewed about their assigned text. They also agree to have their voice cloned and let their words be used and manipulated by the platform’s AI. That way, when a reader pauses to ask the text a question, the text—or rather a language model trained on the ideas of the expert and speaking with their voice—can respond.
Experts recruited to comment on works, besides Kaag and Martin themselves, include: John Banville (Dubliners), Lena Dunham (A Room With a View), Roxane Gay (The Age of Innocence), Marlon James (Adventures of Huckleberry Finn), Laura Kipnis (Romeo and Juliet), Bill McKibben (selections from John Muir), Margaret Atwood (A Tale of Two Cities), Elaine Pagels (selections from the New Testament and Secret Gospels), and James Wood (Chekhov), among others.
Kipnis interviewed Dubuque, the businessman developing and funding Rebind, and wrote about the project and her experience with it for Wired. She says:
The innovation, [Dubuque] explained, was making the… commentary “chattable.” It’s designed to “meet the user where they are”… The App isn’t for everyone, Dubuque stressed. College students can use it, obviously, but the target audience is adults, or at least the book-loving subset. (Five million people in the United States are in book clubs, Kaag had mentioned.) “You read all these wonderful books as an undergrad,” Dubuque said, “and then you graduate and you read newsletters.” The big thing a user has to understand is that Rebind is designed to be an active experience. “If you’re not responding to these questions and thinking deeply, you’re not going to have as much fun.” Kaag too had stressed that the more a reader puts of themselves into the chat windows—highlighting and reacting, producing “marginalia”—the more interesting the conversations are going to become… [Dubuque says:] “It’s going to know what you do and don’t understand, which will make your future reading experiences even better.”
Kipnis asked John Banville if he had any reservations about being “cloned” by the AI. He replied:
But I think that is the case already. I have always felt that there is no John Banville. He ceases to exist the moment I stand up from my desk. I don’t know who he is—I find him a very strange creature. My strange dark brother. So, there’s really nobody in there, just this artistic sensibility, creating stuff. I’m already a clone of myself.
The project has just launched in beta, and you can sign up to join a waitlist to try it here.
I, for one, welcome our new classic book discussion partners. Rebind.ai, join America’s 5 million book clubs!
The recent New York Times article made two points I find noteworthy:
“Mr. Martin and Mr. Kaag … hope to give the Rebind treatment to 100 classics, all published before 1928 and therefore in the public domain.”
“Users will have free access during the rollout, with per-book pricing and a subscription model to follow later this year.”
Kurutz, Steven.
“Now You Can Read the Classics With A.I.-Powered Expert Guides.”
The New York Times, June 13, 2024. https://www.nytimes.com/2024/06/13/style/now-you-can-read-the-classics-with-ai-powered-expert-guides.html.
What expert, still academically credible, published and online pre-1928 commentaries on a classic text might be suitable for training a free large language model (LMM) on a particular classic?What developer of a free LMM, suitable for training on both source text and secondary text, will be first to offer a free alternative to Rebind?I’ve joined the Rebind waiting list – currently #4284.
If you follow this waiting list signup link, you may join yourself, bumping me up in line!
Thank you!
Could this venture worsen the problem of AI hallucinations and false responses? Not that it would be more likely than other AI ventures to yield them — that would depend on the underlying model, as I understand it — but that this venture could make it more difficult to recognize the hallucinations and false responses for what they are?
If I’ve just listened to actual-Margaret Atwood deliver commentary, and I ask AI-Margaret Atwood some questions, I’m concerned I may not apply the appropriate level of scrutiny toward the generated results, perhaps over time if not immediately. Indeed, that I become less likely to evaluate them as generated responses at all. (Substitute ‘I’ for students, if you like. That’s where my mind goes, even though the write-up says students aren’t the main audience.)
This is different from cloning a famous voice for GPS narration or even for all-purpose chatbots. Here, we would be getting a mix of actual-Margaret Atwood and AI-Margaret Atwood, inter-spliced between our question-asking. It seems to me that it could become then harder to resist simply believing too much of what AI-Atwood delivers to the us, listener-inquirers. And isn’t the point of learning to learn what’s true?
I think it’s interesting that this particular venture includes both scholarly experts like Elaine Pagels, and non-scholarly “experts” like Lena Dunham, and people in between like Laura Kipnis (a journalism professor with specialization in gender, not a Shakespeare scholar). In some cases, having a conversation with one of these people, we would expect them to be making factual claims that are well-supported. In other cases, we would expect them to be making interesting claims that we should take with an important grain of salt. It seems to me that a lot of the issues we find with “AI hallucinations” are very similar to the issues that arise with ordinary human mis-remembering, confabulation, and highly opinionated takes masquerading as scholarly consensus.
For the purposes of a “home book club”, I don’t think it’s too problematic if someone ends up with a weird interpretation of literary themes in A Room with a View because AI-Lena Dunham gave her weird take on it. It might be more problematic if they came away with an impression about the dates of authorship of books of the Bible because AI-Elaine Pagels made some authoritative-sounding claims about third century manuscrips.