tech
Booksellers suspect AI firms are buying and then destroying rare books
AI firms quietly bulk buying rare books face resistance from booksellers.
TL;DR
- AI companies are reportedly destroying physical books to train their models, leading to concerns about the loss of high-quality, long-form text.
- Non-destructive book-scanning technologies exist, such as Google's patented method and the Internet Archive's careful, manual approach, but AI firms often prioritize speed and cost.
- The Internet Archive employs trained staff like Eliza Zhang, who meticulously scan books page by page, ensuring preservation and accuracy, even for rare and fragile volumes.
- Recent reports suggest AI firms are engaging in bulk purchases of books, prompting rare booksellers to flag these orders as suspicious, potentially indicating an intent to destroy the books.
- Companies like Anthropic have reportedly used codenames for destructive book-scanning, while others like OpenAI and Microsoft are collaborating on initiatives to preserve public-domain books.
- Booksellers, like Kennys in Ireland, are refusing large, suspicious orders, viewing the potential destruction of books for AI training as a threat to their industry and authors' rights.