[Reflection] #41: The Destruction of Books Is a Background Activity

The news that Anthropic would buy millions of books to slice off their spines and scan the contents, before throwing the whole book away, made the rounds worldwide. People are horrified and keep bringing up 1984, Fahrenheit 451. However, for each title, we’re talking abouta single copy sacrificed among others still circulating in the world, not an attempt to make a work disappear by buying up all its copies. That said, for rare or out-of-print titles of which only a few known copies remain, sacrificing “a single copy” can, in practice, amount to destroying one of the last survivors of an edition. But we don’t know whether any rare copies were part of the batch. The nuance is worth stating: this isn’t organized censorship, but it isn’t a completely risk-free gesture for our cultural heritage either. P/s: Anthropic’s book destruction happened because when it wasn’t destroying books by using digital files of books already scanned, it got itself sued.

Without meaning to defend Anthropic, I find it strange that people get so outraged about a single player, when book destruction is a background activity, a permanent one, that never causes a scandal. I know private individuals who destroy their own books to scan them: they always get stones thrown at them, so much so that in the book-digitizing community, these people stay anonymous. To get a quality scanned version, you usually have to slice off the spine and scan in bulk. Someone once sent me the scan of a rare book published in only 1,000 copies, asking me never to reveal their name, because they didn’t want to be singled out for destroying a rare copy. But it’s thanks to that sacrifice (a single book butchered) that we were able to turn it into an ebook distributed on a far wider scale than the original 1,000 copies.

This individual is the tree hiding the forest. The number of books destroyed every day, without even being scanned, is enormous, and nobody talks about it: publishers pulp their unsold stock by the pallet-load, libraries weed their collections to make room, bookshops that close down liquidate or throw away their unsellable stock. None of these players makes headlines, none gets accused of trying to “erase knowledge” or “rewrite history.” We have a polite word for it in the trade: “pulping,” precisely so we don’t have to say “destruction.”

Anthropic, on the other hand, had the bad idea of leaving a written record of its operation, and of doing it out in the open (well, until it came out): whereas the daily pulping by other players remains invisible.

The book has always been considered sacred, but we should de-sacralize it a little. People still get scolded if they dare to dog-ear a page, annotate in the margins, or highlight, as if every copy, every print run, deserved absolute protection. But a book that truly has value survives the test of time.

If we still cling so much to every copy (including mediocre books), it’s tied to the history of the book: it used to be rare and expensive. Written on parchment (calf, sheep, or goat skin, a precious raw material) at a time when every page was copied by hand. So precious that even the size of the legendary Library of Alexandria, long haloed in the myth of the “total knowledge of humanity,” is still debated by historians: estimates range from just a few tens of thousands to several hundreds of thousands of scrolls – roughly the equivalent of the catalog of your local big-box bookstore or my neighborhood media library in Lille. In the 9th century, Photios wrote the Bibliotheca (or Myriobiblon), a collection of summaries and reviews of hundreds of works he had read. Many of those books have since disappeared, and for many ancient authors, the only trace we have left of their work is precisely the account Photios made of it. These works were lost over the centuries, copied or not, deemed useful or not, sometimes simply burned by accident or by war.

There has always been a test of time that decides which books survive. It used to be called fire, war, damp, a shortage of monk-copyists, a king’s disinterest, censorship. Today it’s also called pulping, library weeding, bookshop closures… and now AI.

Communities like Anna’s Archive have set themselves the mission of “safeguarding all the knowledge and culture of humanity,” by aggregating millions of scans from other digital libraries and distributing them as torrents so they exist in multiple copies, all over the world. The idea being precisely that a file duplicated thousands of times becomes almost impossible to make disappear. These platforms operate outside the law, and have been convicted several times for massive copyright infringement.

Faced with this, it’s hard to know who to applaud: one destroys a copy of every book legally (for the benefit of AI that may free humans from bullshit jobs), the other preserves knowledge but illegally. But each keeps control over a small territory: their own library. The copies we choose to keep at home, to reread, to lend, to recommend, the ones we give a sturdier binding, are all copies that escape, on their own tiny scale, the pulper, the weeder, or disappearance. If we can’t save “books” in general, we can at least save our own.

Further reading: Don’t Count on Getting Rid of Books