| There’s no escaping the fact that AI is increasingly intertwined with the books industry: last month, for example, the announcement of entrepreneur Steven Bartlett’s publishing imprint Flight Books came with the promise of an “AI-powered, data-first, growth-focused” strategy. AI won’t actually be writing the books (that job will be for Bartlett’s choice of celebrity authors, or their ghostwriters), but AI tools will be used to analyse data, and “discover trending topics and points of interest for our authors’ audiences”, according to a spokesperson from Flight. Most of the time when we hear about the impact of AI on literature, it is bad news: authors filing lawsuits against AI companies for feeding their works into the large language models without their permission, for example, or finding AI ripoffs of their books being sold on Amazon. But Harvard literature professor Martin Puchner has been trying to do something positive with the new technology, building customised chatbots so that his students can speak directly with famous figures from history and literature, such as Socrates, Shakespeare and Scheherazade, the storyteller of One Thousand and One Nights. “I built the Scheherazade bot when some of my creative writing students complained that there was no book giving them writing advice based on the One Thousand and One Nights, so I built a bot that extracts writing advice from this story collection,” Puchner says. The academic has always been passionate about making literature more accessible: in the mid-2000s, he edited the Norton Anthology of World Literature, which is used to introduce university students to classic texts (I think I might actually have a copy from my own university days knocking around somewhere). The kind of chatbots Puchner has built “can bring philosophers and writers to life, providing new forms of access and interaction,” he thinks. Puchner himself has enjoyed interacting with his Buddha chatbot: once he asked it “was I talking to the Buddha himself or perhaps an avatar of the Buddha?” “It answered no, it wasn’t the Buddha nor an avatar, explaining that an avatar was the incarnation of a god in human or animal form, but that I should think of it as a bridge between me and the Buddha. Not a bad answer, I thought.” One user of Puchner’s Montaigne chatbot fed back to the professor: “I spent some time with Montaigne, musing on some things that have been bothering me lately and found the time better spent than the many expensive hours of therapy that I’ve occasionally resorted to.” Yet the existence of a computer-generated agent that can write is both “an amazing and confusing thing to happen for anyone who cares about literature,” Puchner says. “All new writing technologies – the alphabet, paper, print, the internet – have transformed literature, so yes, AI will have an impact, though it’s too soon to say what that impact will be.” The professor is in the process of developing a writing course that will incorporate AI where it is useful (for example, for “reverse outlining”, a process of summarising paragraphs into phrases to help make sense of a draft, or “as a sparring partner”) but says it will also emphasise “those aspects of writing that AI is bad at, such as coming up with striking new metaphors and unexpected, improbable formulations and thoughts”. Puchner thinks it will be “intriguing” to see how writers use AI in the research and writing process. It can be easy to get “freaked out” when thinking about that, he admits, but points out that “for most of the 5,000 years of writing, written texts were produced collectively and anonymously by scribes and editors. We might be going back to that mode of production.” | 