As a child, I hated reading. I would dig into pages but the sentences would never stick in my head. At least, until my mother, a brave, 5-foot-tall woman who was a substitute English teacher in the eighth grade, found my report card buried in my school bag. I won't tell you what my reading and writing grades were, but they weren't what she expected.
“You're going to read with me,” she said, pausing to compose herself, “now.”
And that is exactly what happened. Every night after dinner, we sat down at the kitchen table and read together. This was not a story time for me to read on my own, nor an opportunity to listen to a reading aloud; rather, it was my first experience being guided by an expert through difficult books like Machiavelli's The Prince, Salinger's The Catcher in the Rye, and Plato's The Republic. Every few pages, my mother would stop me and ask questions that challenged my thinking about the text, drawing me deeper into the subject and creating space for me to ask my own questions about the author, the text, and its place in literary history. And so I developed a love of reading and became a writer and a philosophy professor.
Most of my students don’t have mothers like mine. I know how lucky I am. And I know that many of my students hate reading. I try to substitute as much as I can, but the mid-sized state university I attend is not Oxford. At a university like mine, which relies on tuition and large class sizes, a one-on-one tutorial format is not possible. But in a conversation I had recently with the director of the Museum of Fine Arts, Boston, we were talking about artificial intelligence and how it can make museums and those little placards next to paintings really dynamic and interactive. I thought about the “Hello Vincent” chatbot that was so successful at the Musée d’Orsay. And, because I’m both a reader and a teacher of classical literature, I thought about books.
Artificial intelligence is about to do for reading what museums already do for exhibitions. We are entering a new era for books, as revolutionary as the invention of Gutenberg's printing press. As I write this, readers already have the power to ask questions of books, and books can answer them.
When ChatGPT took the world by storm in November 2022, my students were the fastest to learn. They loved it. David Smith, Professor of Biological Science Education at Sheffield Hallam University, summed it up nicely: “They don’t want ChatGPT to be badmouthed. They want to be told how to use it.”
But what's interesting is that for students who use it not to produce zombie papers, but to create mock conversations, the best evidence suggests that it enhances their awareness, curiosity, and interest, rather than suppressing them.
When I asked students to create animated conversations with a chatbot on a topic of their choice and then submit the conversations as drafts alongside their finished papers, I found that many students produced work that was far more engaged and polished than anything they had ever produced before. And, more importantly, they were excited about their work; they enjoyed the process of writing it. A recent study at the Massachusetts Institute of Technology backs up my experience: when students engaged with a chatbot as part of their creative writing—when they used the chatbot as one of many tools in their toolbox for producing original writing—they learned faster, better, and more satisfied.
What if we could embed an AI-enabled chat window into a text, providing a place for real-time conversation with someone as wise as my mother? What if one of the classics of literature or philosophy was accompanied by commentary by experts who have studied the text for years, and that commentary was delivered directly by an artificial intelligence acting as an interlocutor for the reader? In this way, we could generate conversations about the book using the Socratic teaching method employed not only by my mother but by almost every great thinker in history. We could generate the same excitement and enhanced learning of classic texts that our students show when creatively interacting with chatbots.
I teamed up with researchers based in Boston, California, and Ukraine to dictate 30 hours of responses to nearly 1,200 questions about Henry David Thoreau's experimental memoir, Walden, as well as 1,000 questions that might pique readers' interest about the text. With this material acting as a chatbot's contextual window into the text, readers could ask almost anything and receive answers that were surprisingly similar to my own, with follow-up questions I crafted that provided readers with unexpected and meaningful depth into Walden.
In our tests, we quickly proved that the conversations around the books we were working on were much more dynamic than we had imagined. The ability to train Large Scale Language Models (LLMs) on literature written by compelling scholars also means that the LLMs can help bring literature to life for countless readers who might never have come across the book otherwise.
The experience of dictating Walden from scratch was often boring; just talking into a recorder and trying to come up with all the questions my students have asked about it for years. I realized that being interviewed about the book could make the whole experience of transforming into a chatbot not only fun, but meaningful, and perhaps most importantly (this is why we love the classroom), dynamic, exciting, and fun. So at the end of my Walden experiment, I enlisted my friend, educator, and Thoreau expert Michael Goodwin, to ask me some final questions about Thoreau's classic work.
My prediction was correct. Michael, playing the inquisitive student, elicited the most personal and thought-provoking content—insights I didn’t even know about, much less express myself. And so an important idea was born for this new era of AI reading: if AI platforms use real conversations about meaningful subjects as their training grounds, it will be more likely that users can use them to have rich simulated conversations about meaningful subjects. I tested this hypothesis this spring by interviewing Booker Prize winner John Banville about James Joyce’s Dubliners, and the results were surprising. The conversations on the e-reader resembled Banville’s conversations with me, and the responses were similar.
Can it pass the Turing test and sound like a real human? Maybe it can, maybe it can't. It depends on the question you ask. (If you treat it like a human, it will sound eerily real, but if you try to fool it, you still can.) Either way, it was in a whole different class from ChatGPT's cardboard-box-like answer. Nobody wants to talk about great books with ChatGPT, but people care very much about opinions about classics from people they respect. Trust me. Generative AI will be a tool to deliver the author's authentic insights and perspectives to readers in unexpected ways.
Yes, I have goals. I want everyone to love reading as much as I enjoy it now. And I want all of us to use AI in ways that aren't pointless, disappointing, or just plain annoying. At heart, I'm a teacher, as is my mother, and I believe this is (pardon the expression) one of the great educational moments of our time. And, like the invention of the internet itself, it's also a rare time when knowledge is being radically democratized for the better for all.
Before Johannes Gutenberg invented the printing press in 1440, the written word was buried in handwritten manuscripts, read by a select few. By the 17th century, literacy rates were soaring and the written word was beginning to spread. But many of the most difficult yet valuable classics, like Joyce's Ulysses and Plato's Republic, remained inaccessible and incomprehensible, despite the fact that they could only be understood by careful study under gifted teachers—and only to a select few.
The elitism inherent in certain educations has remained almost unchanged for years. Soon, in fact, it is happening right now, we will be able to expand our simulated and interactive conversations on a vast number of classical works from all traditions: literary, scientific, philosophical, and spiritual. The birth of the Master of Laws and its possible integration into the process of close reading will revive and revitalize dying words, ancient perspectives, and even dead languages. Yesterday, I used AI to translate a text from Sanskrit and Pali to Tibetan, Koine, and Sumerian. I was direct messaging a friend in Mexico about this, with amazement and delight, when in reply he asked me, “What about Nahuatl (the language of the Aztecs)?” I learned that the Sanskrit word for the enlightened mind (‘bodhi’) in the ancient Aztec language is “Tlamatiliztli.”
I believe that within months, not years, we will see the emergence of useful reading tools to help students break down barriers to comprehension. The commentary I am describing for Walden (which I think of as a “rebinding”) is based on a deep belief that reading is a personal and interpretive experience. Commentators who develop this new type of interactive book can model and inspire the personal close reading behaviors we want our students to learn. And AI will be used to amplify and expand the message of these teachers.
Will “real” teachers be replaced by AI? No. Students spend most of their time reading assigned books outside of class, not in the classroom. Should students use AI to have better, deeper reading experiences outside of the classroom? Absolutely! The more prepared students are to discuss texts, the more everyone will learn. I think teachers will not have to spend so much time learning the basics, and discussions can reach a much higher level of engagement, more frequently, with more students.
Teachers, too, will embrace AI-enhanced reading experiences, preparing them as best they can for the magic of live, human group discussions. Nearly 20 years ago, I was asked to teach Walden for the first time. I wanted a little help. A chance to ask an accomplished teacher how they did it. So these interactive books are not just useful for students. The academic world will quickly wake up to the fact that the LLM is a general-purpose technology (similar to an electric motor). There are all kinds of creative and authentic ways to leverage the LLM, and we're just beginning to explore it.
Finally, I would like to say a word about the ongoing lawsuit against OpenAI being brought by various prominent authors and other intellectuals. One thing is clear: this lawsuit is not like Napster or Spotify threatening to steal content from the music industry. OpenAI is not taking a corpus of books to resell.
The fact that we are using literary works to teach our LLM should get us thinking, and copyright issues make sense. But instead of worrying about who owns what, let’s think about the bigger picture for a moment. What does bringing world knowledge into our LLM actually mean for us? It means that in the very near future, you will be able to converse with the books you are reading. And it is up to teachers like us to get ahead of these technological trends and, if not protect them, at least provide the expertise to help students in their inevitable conversations in the virtual world.
Socrates warned us that the written word is just as dangerous as a painting, because words cannot refute or answer themselves. But now they can, and if we are careful, they can refute and answer in the very ways that Socrates recommended for the pursuit of truth.
John Karg produced Thoreau's Walden for the e-book platform Rebind, which he co-founded, due to be released in August. He is the Donahue Professor of Art at the University of Massachusetts Lowell and an external professor at the Santa Fe Institute.
