From the Opinions Editor: Anthropic is destroying books to make data. For the reader, an existential challenge

The Italian miller lived in the early decades of the Print Revolution, when books were beginning to escape the cloisters of monasteries, courts and universities and enter the lives of people far removed from traditional centres of learning. In all likelihood, he was doing what generations of book lovers after him have done: Talking to books, arguing with their authors, fitting the written word to their own experiences, even misunderstanding texts and arriving at conclusions entirely unanticipated by the creator of the original work. Every generation after Menocchio has thrown up new heretics. The profusion of fatwas, injunctions decrees and hurt feelings are reminders that books have rarely been innocent objects in the eyes of those who guard orthodoxy. But today, the relationship between reader and writer faces a new kind of challenge.

Reports that AI companies have acquired and destroyed thousands of printed books after digitising their contents for machine learning point to a significant change in the economy of knowledge. The book risks becoming less a work to be read than a vessel from which information can be extracted. Once the text has been captured, the physical object can appear to have served its purpose.

There is something deeply unsettling about this idea. The bibliophile is perhaps like the music lover: Like listening to an album rather than an assortment of individual songs, reading can mean experiencing a piece of text in its entirety, pausing or rereading before returning to favourite sections time and again. Like a loved prelude or interlude, the felicity of a sentence or even a phrase can be striking. The marks left by the reader — a bookmarked page, a note scribbled in the margin, an underlined sentence — are as precious as the printed word.

AI, in contrast, can extract the words from a million books in a fraction of the time it would take humans to read them. It can identify patterns across texts, make connections and generate new combinations of knowledge. In Anthropic’s Project Panama, the digitised books dissolve into statistical memory, where language and knowledge are abstracted and reassembled into machine-generated responses. What, then, happens to the social role of the printed world when the book is reduced to the information it contains?

Menocchio was disruptive not merely because of what he believed, but because he claimed the right to make meaning for himself. Religious authorities worried about readers interpreting scripture for themselves; political authorities feared pamphlets that could spread dissent; colonial governments censored books because ideas travelled with them. The novel, newspaper and cheap paperback each widened the circle of public debate. They passed from hand to hand and generation to generation, creating communities of readers who may never meet and conversations across centuries and cultures. Libraries, bookshops, lending circles and classrooms have grown around this shared world of thought.

AI, too, can widen access to the world of knowledge, allowing a student far from a major university to ask questions, translate texts, search archives and encounter specialised knowledge with unprecedented ease. But it also changes the social architecture through which knowledge is encountered. Instead of entering a library or leafing through a book and following an author’s argument, the user often receives an answer assembled from many sources. Anthropic destroyed millions of books to perhaps create a machine that can reproduce or manipulate their contents. But what about the sources, arguments and intellectual traditions behind this corpus?

What then will become of the unpredictable encounter, the kind that created Menocchio’s cosmology? Will readers still find the friction that makes them question what they already believe, or the strange idea that sends them down an intellectual path its author never intended? Books — printed or digital — are unlikely to disappear. What will, however, happen to reading when the efficient retrieval of information becomes its predominant function? Will there still be enough unruly readers like Menocchio?

Till next time, I leave you with these questions.

Stay well
Kaushik Das Gupta

Recommended Readings:

Christopher Nolan’s ‘Odyssey’ is spectacular. It just doesn’t get its hero right

Best of Both Sides: Indian cinema isn’t struggling, it’s just a Bollywood problem

Best of Both Sides: For Bollywood’s woes, blame the ecosystem, not the industry

Message from the young: We don’t feel at home

A K-pop band’s plea to fans: Put down the phone