India, Oct. 4 -- At first glance, libraries haven't changed much. They still have shelves, reading nooks and that familiar quietude of heads buried deep in pages, but with a few significant plot twists. Robots trundle between rows of books to deliver readers their requested titles. Artificial intelligence-guided "librarians" answer questions or offer recommendations, while invisible sensors quietly adjust the lighting and the temperature of reading nooks. Versions of this sci-fi dream are already playing out across the world. l Visitors walking into a reading room in Changning District Library in Shanghai, China are greeted by Xin Ye, a digital human avatar on a screen. The AI-powered virtual assistant can scan members' faces or library cards; sift through their borrowing records to pick out books they'd possibly enjoy; book a suitable seat; have a robot deliver titles to their seats. l In the country's capital, the 75,000-sqm Beijing Library (below) is fashioned like a giant ginkgo tree - a 290 million-year-old species native to China. Its automated storage and retrieval system can plod through its collection of over 8 million books to deliver a reader's chosen title in less than 15 minutes. l Meanwhile, book-delivering robots have roamed the halls of the National Library of Public Information (NLPI) in Taiwan for about a decade. In 2023, the country's first generative-AI-guided "librarian" stepped in to help the staff. Called Xiao-Shu (meaning well-versed in Chinese), it appears on a screen in human avatar and offers readers consultation, book recommendations and sometimes just a chat. l At Singapore's National Library Board-operated libraries, shelf-reading robots (right) have been helping the staff solve a notoriously familiar problem: spotting misplaced books. Every night, the robot paces up and down to scan books - equipped with RFID (Radio-Frequency Identification) chips - to make sure they're in the right place and order. By morning, it delivers a report to the library staff about titles that are out of place, saving up to 3,500 staff hours per year. Across the world, libraries are gradually adopting or experimenting with AI, machine language, national language programmes and other technological interventions to make storing, cataloguing, curating collections, as well as borrowing and reading easier. What does that mean for the idea of this institution that has traditionally served not just as a repository of knowledge, but also a safe space for privacy, plurality, democracy and discovery? We'll come to that in a bit. First, a quick recap. Across millennia, librarians have confronted essentially the same problem: how does one make the overwhelming amount of human knowledge stored in books easily accessible ? At the oldest known library, in the royal palace of the Assyrian king, Ashurbanipal, in 7th-century BCE Iraq, some of the earliest attempts to answer that question began. Administrative accounts, mathematical, medical and literary texts were organised and preserved on clay tablet catalogues. Over time, these catalogues shapeshifted to include papyrus scrolls by the 3rd century BCE in Egypt, and then codex - handwritten and handbound stacked sheets of paper - in the Middle Ages. As books became cheaper and more plentiful, libraries needed better ways to keep track of rapidly growing collections. Card catalogues, reflecting the name of the book, author, subject and number emerged by functioning as a sort of search engine. By the '60s and '70s, computers began speeding things up. Electronic catalogues eventually replaced drawer cards altogether. Readers could search thousands of titles in seconds by author, subject or keyword, and later do so from home, too via library apps or websites. Barcodes and RFID, meanwhile, accelerated the process of borrowing, returns and stock-taking. Now, the more interesting conundrum centres around what libraries are asking AI to optimise. Is it workflow, organisation, storage and retrieval, or something more human, such as recommendation? Algorithm-driven recommendation systems are good at giving us more of what we already like. However, a library also introduces us to books we never knew we wanted to read. After all, how often do we walk into the library looking for a cosy dramatic read, only to leave, after a librarian's recommendation, with a thrilling piece of historical fiction, or poetry? If libraries made information accessible, the internet, and now generative AI, has made it abundant. The challenge at hand is identifying what to trust. Libraries will continue to play a central role in media and digital literacy, helping readers separate fact from fiction, spotting loopholes and accurately tracing a claim back to its source, says Lotta Muurinen, director of regional library services, City of Helsinki, Finland. They will also function as a safe haven from digital spaces, adds Noora Hirvonen, professor, information studies, University of Oulu, Finland. This is essential as we increasingly get locked in our own bubbles, where we seek and receive information that confirms our beliefs, no matter what they are. Retaining the joy Getting that balance right will require guardrails, especially around privacy. Libraries have traditionally offered considerable privacy around what somebody chooses to read. On the other hand, the more the algorithm knows about your searches, borrowing history and behaviour, the better it can predict what you want next, says Kashyap Kompella, a tech industry analyst and author. Data privacy is one of the major reasons why libraries across Finland are still evaluating integrating AI into their workflow, even as others around the world are experimenting with it. Then there is the question of how much efficiency is too much efficiency? Searching a database is faster than wandering down rows of books. An AI summary, prompted by just a word or two, is even faster. But speed and efficiency can come at the cost of something more valuable.The unexpected book waiting on the next shelf. A sudden conversation that shifts perspectives. A rare space to simply be....