Industry & Business Hacker News (AI)

AirTag reveals Amazon is trashing rare books to train AI

AmazonAI training datarare booksinvestigation

Booksellers have long suspected that AI firms were buying up huge lots of rare books and destroying them after scanning, but hard proof was elusive. Now, 404 Media reports that an AirTag planted by a bookseller in a rare book tracked that book to an Amazon AI training facility in Las Vegas. There, a team focuses on tearing books from their spines and scanning pages, working out of a warehouse with the logo VGT3. A T. rex preparing to devour a book adorns the door, a tone-deaf touch given the escalating backlash over destructive book scanning.

Amazon declined to comment on the specific findings, offering only a generic statement that it purchases books through commercial channels to develop and improve its products and services. However, the report connects this practice to Amazon's development of frontier AI models, which require massive amounts of unique training data to stay competitive with Google, OpenAI, and Anthropic. Since rivals carefully guard their training data, rare books provide a distinct advantage—Anthropic and xAI have publicly stated they are not training on rare or antique books. Forum posts from VGT3 workers suggest Amazon had been running low on books to scan, at one point exhausting supply and fearing the warehouse might shut down, though it remains operational.

The confirmed destruction of rare books has horrified many book lovers—one critic, Michael Burry, reportedly called the practice 'evil incarnate.' Meanwhile, Amazon workers described the gig as a 'nice' opportunity for those seeking a humdrum job with flexible hours. 404 Media also suggested that the investigation helped firm up another bookseller theory about why AI firms might be buying rare books, though the excerpt cuts off before providing details.

Read original →

← Back to home