Amazon is reportedly purchasing vast quantities of rare books, cutting off their spines and scanning them to train artificial intelligence models. According to 404 Media, which placed a tracking device in one such book, the rare volumes eventually reached an Amazon facility in Las Vegas marked with a dinosaur clutching a tome.

The company claims it is buying books through commercial channels to improve its products and services for customers. However, these rare texts offer unique training data, especially those out of print or not available online, feeding hungry LLMs that risk model collapse from overconsumption of AI-generated content.

Companies like Amazon need endless volumes of text for their advanced language models, already digesting what they can from the internet and beyond. This new source of training material has raised concerns about the preservation of cultural heritage in the digital age.

The irony is not lost on many: as we strive to make our digital assistants more human-like, are we losing touch with the very history that defines us? Will future generations wonder what stories lie behind these spines?