After Spanish physician and theologian Michael Servetus was burnt at the stake in 1553 for religious heresy, only three surviving copies of his The Restoration of Christianity would be spared from the bonfire. One particular copy, now held by the University of Edinburgh library, is ironically believed to have been personally saved by the Protestant leader John Calvin, who was responsible for Servetus’s execution in Geneva. Anyone familiar with Calvin, who bragged in a letter that he had “purged the Church of so pernicious a monster,” knows that the book wasn’t preserved due to affection for Servetus, but rather a bibliophile’s sentiment — even a doctrinaire, zealous, tyrannical bibliophile. And yet, as bizarre as Calvin’s motivations might have been, he’s still not as noxious as the artificial intelligence companies, including Anthropic, that have been involved in bulk purchasing, digitizing, and then pulping out-of-print books to train their large language models. Calvin at least had the excuse of his God; AI companies don’t even have that.
In a sobering report by Frank Landymore in Futurism, published last Saturday, explains that Large Language Models (LLMs) are increasingly trained on their own slop, creating a need to train them on pre-2022 writing that is uncontaminated by AI itself — that’s the particular sales pitch made by ISBNdb, a company supplying bulk orders of books to tech companies — with purchases of between one thousand and one million volumes at a time. A “hydraulic powered cutting machine” slices off book spines and removes the pages, which are scanned and digitized before being pulped. This process of “destructive scanning” is a way to elude copyright regulations through a legal loophole, with the result being that “an AI company could be destroying some of the few remaining copies that can be found” of these books.
The destruction of books, as Servetus’s case demonstrates, is hardly a historical rarity. Nor is the pulping of books for non-ideological reasons uncommon, as publishers are responsible for destroying nearly 25% of works released each year, almost always because they didn’t sell.
Rather, what the recent revelation portends is unique and terrifying: the rapacious, cannibalistic devouring of culture so as to feed the very technology meant to replace the human. The indifference of the LLM is chilling in its own way; one imagines that if a copy of Servetus’s book fell beneath the lifeless eye of the digitizer, it would simply upload all of those theological arguments into the hungry maw of the machine, where it would only serve as fodder for the advanced plagiarism device which is artificial intelligence.
“We see a future where intelligence is a utility,” said OpenAI CEO Sam Altman, “and people buy it from us on a meter.”
This is a future where technocracy has accomplished its war against the sacred materiality and artistry of the book itself, where everything is reduced to so much predictive binary. It’s an act of literal erasure for which George Orwell’s cliched warning that “who controls the past controls the future” is restored to its original impact. LLMs swell while the shelves starve, the result being an act of libricide, of corporate indifference to the obliteration of culture, ideas, and humanity itself.
