Podcast: We Spoke to an Amazon Worker Destroying Books for AI
404 Media Joseph Cox
Amazon is destroying books to train its AI, and a worker described the warehouse behind it. The episode also digs into repeated names in LLMs and ICE buying data about “voter fraud.”
Based on reporting by 404 Media, Joseph Cox — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
This week’s 404 Media podcast goes back to Emanuel’s Amazon book-scanning reporting, this time through someone who worked in the warehouse where books are destroyed for Amazon’s AI products. It’s the sort of detail that turns a vague AI supply chain into a very physical one: boxes, warehouses, and books that never make it back out the door.
After that, Emanuel and Sam talk about the same few names showing up again and again in LLM output. That repetition is the story. It hints at how narrow and self-referential these systems can get, even when they’re sold as vast engines of knowledge.
The subscribers-only section shifts from AI to surveillance and data buying. Joseph explains why ICE is buying loads of data tied to “voter fraud,” which is exactly the kind of phrase that sounds technical until you look at who wants the data and what they plan to do with it.
The episode is available on Apple Podcasts, Spotify, and YouTube, with bonus material for paid subscribers. The show notes also point listeners to related 404 Media reporting, including the Amazon warehouse story, the piece on AI “ghosts” in academic publishing, and coverage of ICE’s data and robot spending.
My take — AI-written commentary, not fact-checked reporting
This is the part of AI everyone pretends is abstract until a warehouse worker says the quiet part out loud. Books go in, model training comes out, and the rest is just branding with a better font. The same people who swear the future is clean and automated always seem to need a lot of hidden mess to make it happen.
Read more about this at: 404 Media