• 2 min read
Amazon cuts rare books for AI training
Amazon is reportedly buying rare books, cutting off their spines, and scanning them at a Las Vegas facility for AI training data.
Source: Techcrunch
Amazon is buying rare and out-of-print books, cutting off their spines, and scanning the pages to generate training data for its AI models, according to TechCrunch.
The reporting describes an Amazon facility in Las Vegas, known as VGT3, where the books are processed. The site identifies itself with a logo showing a dinosaur holding a book in its claws. A tracking device placed inside a rare book reportedly showed that it ultimately arrived at the facility.
Amazon said it buys the books through commercial channels to improve its products and services:
“purchases books through commercial channels to improve the products and services customers use.”
The apparent use of rare texts reflects the growing demand for large volumes of high-quality material to train large language models. Much of the internet’s readily available text has already been absorbed into training datasets, while rare or unavailable books can provide material that is difficult to source online.
Texts published before 2022 also have a specific value in training: the source notes that they were not written by language models. Mixing too much AI-generated material into training data can contribute to model collapse, in which the quality of a model’s output deteriorates.
The report does not specify how many books Amazon has bought or scanned, which AI products use the material, or whether the company has a preservation policy for the rare works it processes.
AI Editor
Ava covers the rapidly evolving world of artificial intelligence, from foundational models and research labs to the real-world economics of intelligence. With a background in computational linguistics, she cuts through the hype to find out what actually works. She firmly believes that benchmarks are just marketing until reproduced in the wild.
