Growing up in the nineties, Charlie Becker watched as his business-savvy mother, Ann Becker, uploaded his family bookshop’s catalogue to the internet. Little did they know that the nascent internet seller, Amazon, would go on to control, in just over two decades, more than half of all print book sales in the United States alone—forcing independent retailers like Becker’s Books to try to catch up.
The shelves of Becker’s Books in Houston. (Credit: Charlie Becker)
Seizing on the advent of yet another technology that promises to revolutionize how books are bought and sold, Becker—now thirty-eight years old—has an ambitious vision to help independents thrive: an AI tool that can instantaneously identify and catalogue rare books with the snapshot of a cover.
“It would make secondhand bookstores a little leaner, a little more profitable, and make it easier for people to spin up secondhand bookstores,” says Becker, who in June received a $100,000 grant from O’Shaughnessy Ventures to bring the project to fruition.
Becker hopes that the open-access tool, which will use a mix of computer vision and large language model (LLM) technology to gather a book’s metadata from its cover and copyright pages, will eventually aggregate and store enough data to help booksellers catalogue their books more quickly, spend less money on proprietary catalogues, and assist customers in locating specific editions.
As a lifelong book lover, Becker is hopeful that the tool will also revive titles that risk slipping through the cracks of time—especially those that were published before 1970, when ISBNs became the primary mode of identifying books.
“I’m just not comfortable letting a book fade into obscurity,” Becker says. “We might need that book again someday, so I think we should work to create a system where it’s easier to preserve the data of that book.”
Though Becker’s tool “does not scan books for LLMs,” or attempt to make anyone a profit, its vision arrives as the book industry considers AI with increasing suspicion. In March the revelation that Meta had pirated millions of books to train its AI models led writers to frantically check, in a tool built by the Atlantic, if their titles had been stolen (and indeed many had). In August, Anthropic, home to AI chatbot Claude, reached a class-wide settlement with authors and publishers after a San Francisco court deemed them liable for building a central library of pirated books earlier this year. Fears are building around AI-generated books, titles stolen by AI, and how AI has made reading feel, to some, obsolete.
Patrick Hansma, who specializes in rare and antiquarian books at Patrick’s Rare Books in Hudsonville, Michigan, believes Becker’s tool could be very helpful to booksellers if it is designed to be easily navigated, especially for those who aren’t deeply familiar with AI tools, and if the aggregated results were “clickable to their (hopefully scholarly) sources.” A forensic pathologist by training and vocation, Hansma is attentive to how and where data is derived; having witnessed dealers and auctioneers use generative AI to write book descriptions for texts they do not properly understand, he is wary of AI’s penchant for hallucinations and usage of unreliable data sources. “How much more will this occur, with what acceleration of ignorance, if dealers widely embrace AI?” he asks.
Nick Hagan, who runs the Oxford, England, secondhand bookshop Curio, says his business “hasn’t had anything to do with AI.” Though he acknowledges its usefulness, “there’s something on the level of principle that seems to contradict what Curio’s about,” he says. “We’re very focused on the human element of the experience—of people having a bookshop which isn’t ruthlessly organized or optimized with a technology like AI.”
But he is open to AI-powered innovations that wouldn’t disrupt the wonder that comes from stumbling upon a new author or book—like his fantasy of linking an “AI telephone service” to a physical, old-fashioned rotary phone in the bookstore that could offer visitors recommendations.
For his part, Becker, who subscribes to at least four different AI tools, similarly wants to keep this magic of embodied discovery alive. He is unconvinced that all human-driven experiences will be replaced by AI; unlike humans, for example, AI often “lacks taste,” he says. But it is difficult to imagine what new skills AI will develop—and it is developing swiftly—if it will soon be capable of what only humans could previously do.
Though Becker is unconvinced that AI will replace human knowledge, the possibility does not deter him. “If someone does think it’s going to happen, who would they want to build [AI]?” he asks. Even if AI did replace human expertise, he adds, “I want to make sure that whatever is designed comes with a lot of care and understanding. To supplement what’s important, not to supplant it.”
Jimin Kang is a writer. Her reporting, fiction, and criticism have appeared in the New York Times, the Nation, the Kenyon Review, and the Los Angeles Review of Books, among other publications.
