A media survey indicated that Amazon was supplementing AI training data through the procurement of paper books. After being sent to a facility in Las Vegas, the books were removed from their ridges, scanned page by page and subsequently destroyed in paper form.
AirTag tracked down the Vegas warehouse.
Independent technology media 404 Media placed Apple AirTag in a rare number of book orders, followed up and found that the cargo eventually reached the VGT3 warehouse in Las Vegas. The report cites staff statements that the main task of the team was to receive a large number of books and to remove binding parts so that pages could be sent into the scanners more quickly.
By 2022, books were more popular with AI.
Generating AI recently re-purchased a large number of paper books, one reason being that the content published before 2022 was less contaminated by machine-generated text and that many texts were not fully online. For model developers, such materials are more suitable for use as raw training language.
The report mentions that if models repeatedly absorb content generated by AI itself, output quality may continue to decline. As a result, earlier published paper books that have not been digitized on a large scale are becoming scarce resources in the training data market.
Book scans are warming up in tandem with copyright litigation.
Reports indicate that the Amazon is not an example. A supply chain dedicated to procurement, dismantling, scanning and discarding paper books has emerged around AI training data. Previously, Anthropic's internal project was exposed to a similar large-scale digitization of millions of books.
Paper books for the legality of AI training have recently entered the jurisdiction of United States courts. It was mentioned that a federal judge had previously ruled that the use of legally purchased book training AI might constitute a reasonable use. However, cases involving pirated scanning are not the same. Anthropic had previously reached a settlement of $1.5 billion in disputes relating to the scanning of pirated books, and Salesforce was facing a class action for alleged theft of books.
Additional information:According to the Amazon response, companies “buy books commercially” to help develop and improve customer-oriented products and services. According to Media, this was the first time that a relevant operational location had been publicly directed to a specific facility.