Amazon buying rare, ‘human-written’ books to train its AI model: Report

A representative picture.Amazon is buying rare books, cutting off their spines and scanning them to train its AI models, according to 404 Media.The technology giants, like Amazon and Anthropic (which has ingested illegally pirated books), are using these rare, out-of-print books as a new source of training data, since these texts are impossible to find on the internet and are guaranteed to be human-written.According to the report, Amazon is carrying out this activity at its VGT3 facility in Las Vegas, which identifies itself with a symbol of a dinosaur holding a book in its claws.These texts are particularly valuable because anything published before 2022 is guaranteed to be human-written.

Training on AI-generated text risks "model collapse," where an LLM's output quality degrades after ingesting too much machine-written content.Amazon told 404 Media that it "purchases books through commercial channels to improve the products and services customers use”.