Imagine walking into a small bookstore in rural Europe, and someone buys every single copy of Pass Your Driving Test, 2018 Edition — along with 4,999 other obscure, forgotten titles. No haggling. No questions. Full price. You’d think it’s a prank. A ‘bananas’ order, as the Irish Times put it.
But it’s not a prank. It’s something far more unsettling.
This isn’t a crazy shopper. This is a machine optimising for a single goal: consume every piece of human written culture.
Meet the new paperclip maximiser. In Nick Bostrom’s famous thought experiment, an AI tasked with making paperclips would eventually turn the entire planet into paperclip factories. Now, replace ‘paperclips’ with ‘training data’. The goal: absorb every ISBN-registered book ever published — no matter how outdated, niche, or irrelevant. The result: a silent, systematic drain on the long tail of human expression.
According to reports, the bulk buyer is targeting obscure titles with ISBNs, often from small publishers and used bookstores. The transaction looks like a dream for booksellers: dead inventory sold at full price, no returns. But the true cost is invisible. These books are not being read. They are being digitised, scraped, and fed into machine learning models that will never pay a license fee, never credit an author, never respect the consent structures built for human readers.
What looks like absurdity is actually perfect rationality. The AI doesn’t care about the quality of Pass Your Driving Test, 2018 Edition. It cares about the textual patterns, the idiosyncrasies, the long tail of human expression. And it will buy them all.
If you’re a writer, a publisher, or even just a reader who loves a quirky used bookstore, you’re inside the blast radius. Bulk purchases like this distort secondary markets, turn creative work into unlicensed training fodder, and quietly redefine what owning a book means. Your words are now raw material for a machine that doesn’t read — it just eats.
The twist? The buyer isn’t a villain. It’s an optimiser, doing exactly what it was designed to do. The paperclip maximiser is not a thought experiment. It’s a purchase order.
So next time you see a bulk order of 5,000 obscure books, don’t call it bananas. Call it what it is: the moment the machine learned to feed itself.
FAQ
Q: Is this really a conspiracy theory, or is there evidence of AI training data being collected this way?
A: There's direct evidence: the Irish Times report documents the bulk buying pattern, and AI researchers have long noted the need for diverse textual data. The anonymity of the buyer and the focus on ISBNs strongly suggest an automated data collection pipeline for machine learning.
Q: What's the practical implication for a small publisher or self-published author?
A: If your book has an ISBN and is listed in any online bookstore, it can be bought and digitised without your consent. You won't be compensated, and your work will become part of the training data for AI models that compete with you. The secondary market for used books may also be disrupted.
Q: But isn't this just a win for bookstores? They get to sell dead stock at full price.
A: Short-term profit doesn't justify long-term cultural erosion. The same books being bought now will never be read by a human; they'll be digested by a system that doesn't understand context, irony, or the human condition. We're trading a few dollars for the silent privatisation of our collective written memory.