AI Companies Are Buying Antique Books, Ingesting Their Contents to Train Models, and Then Destroying Them at Incredible Scale

AI companies are bulk-buying physical books published before 2022, destroying them to scan and ingest their contents for training language models. This practice exploits first-sale doctrine and targets pre-LLM texts as 'clean' training data, with intermediaries like ISBNdb facilitating anonymous bulk purchases while acknowledging the poor optics of destroying rare and out-of-print books at scale.

Teaching:
• Use the physicality of practice as a hedge against digital contamination—just as pre-2022 books are valued for being AI-free, embodied practice offers knowledge that can't be extracted or replicated by text alone
• Frame the destruction of books as a cautionary tale about extraction versus stewardship: practice isn't about consuming sequences for output but about tending to the lineage with care
• Teach students that attention is a finite resource like rare books—once you let it be bulk-processed by distraction or automation, you lose the uncommon, irreplaceable thing
• Draw parallels between 'transformative use' legal loopholes and spiritual bypassing: both claim transformation while actually just repackaging and destroying the source

Writing seeds:
• Essay on practice as anti-extraction: why the Ashtanga method resists being turned into data or content, and what that teaches us about systems that can't be optimized away
• Short post comparing AI's hunger for pre-contaminated texts to the yoga world's obsession with 'authentic' lineage—both seek purity in the archive while destroying the living thing
• Piece on first-sale doctrine and the myth of ownership in practice: you can buy a book or a spot in a class, but you can't own the transmission; transformation isn't the same as understanding
• Essay titled 'The Optics Problem Is Real' on how yoga marketing sanitizes the uncomfortable truths of practice the way ISBNdb sanitizes book destruction for AI labs

Idea map:
• Connects to systems literacy: AI companies treat books as isolated data points rather than nodes in a living knowledge system, just as students often treat asanas as isolated poses rather than parts of a coherent practice architecture
• Reinforces embodiment theme: physical books are valued precisely because they can't be digitally replicated without destruction; embodied practice similarly resists extraction into text or video
• Extends attention work: the article shows how attention economies commodify and destroy sources; practice trains sustained attention as resistance to this extractive logic
• Deepens practice-as-method: the 'transformative use' defense mirrors how practice transforms the practitioner, but here transformation is used to justify destruction rather than care

Source: https://futurism.com/artificial-intelligence/ai-companies-destroying-rare-books