Amazon’s AI Book Hunt Raises Alarms Over Rare Volumes Destroyed for Data

Amazon’s AI Book Hunt Raises Alarms Over Rare Volumes Destroyed for Data
Amazon’s push for ever more AI training data has collided with a more tangible concern: rare physical books that critics say are being sacrificed to feed the machines.

The scrutiny began after 404 Media reported that AI companies were buying used books in bulk, a market where out-of-print volumes can contain text unavailable online. Amazon’s need for material is especially acute as models exhaust easily accessible web data and face the risk of degraded outputs when trained repeatedly on AI-generated text.

The concern sharpened when 404 Media placed a tracker inside a shipment of rare books and followed it to Amazon’s VGT3 facility in Las Vegas. Employees at the site said they receive huge shipments, remove the bindings to accelerate scanning, and leave the printed books destroyed in the process. Amazon said it “purchases books through commercial channels to improve the products and services customers use.”

That explanation frames the process as ordinary procurement; the reporting frames it as an irreversible loss. The dispute is not over whether books contain valuable data, but whether rare and hard-to-replace objects should be reduced to text files in the race to build better models.

The contrast is particularly stark against the corporate language of empathy often attached to technology leadership. Microsoft chief executive Satya Nadella has argued that “Empathy is not a soft skill,” calling it “the hardest skill we learn.” His comments do not address Amazon’s operation, but they underscore the question hanging over it: whether AI’s expansion can account for values that cannot be scanned, indexed or reproduced.

Continue reading https://foxvector.com/stories/01a024b9-ec8c-18fe-7095-2e176aedab83