Report says Amazon is scanning rare books for AI training

In short: A report says Amazon is buying rare books, cutting them apart, and scanning them to help train AI models.

What happened

TechCrunch reports that 404 Media tracked a shipment of rare books that ended up at an Amazon facility in Las Vegas. 404 Media said it put a tracking device inside a rare book, and the book later arrived at the site.

According to the report, Amazon buys large numbers of rare books, cuts off their spines, and scans the pages. This turns the text into a digital copy that can be fed into an AI model. (An AI model is a computer system that learns patterns from lots of examples, like a student learning by reading many books.)

The facility is reportedly known as VGT3. Amazon told 404 Media that it “purchases books through commercial channels to improve the products and services customers use.”

Why it matters

AI companies need huge amounts of text to train large language models, often called LLMs. These are the systems behind many chatbots and writing tools. Many of them have already learned from much of the text that is easy to find online, so older and out of print books can be especially valuable.

The article also notes a quality risk when AI systems train on text written by other AI systems, sometimes called “model collapse” (like making a photocopy of a photocopy until it gets blurry). That is one reason publishers and libraries worry that rare printed books are being treated as raw material for data, even when the physical copies could have long-term cultural value.

Source: TechCrunch AI