Elon Musk Orders SpaceXAI to Use Non-Destructive Book Scanning for AI Training
Elon Musk announced that he has instructed the SpaceXAI team to use non-destructive scanning methods to digitize rare books for AI training. This directive comes in response to reports that AI companies, including Anthropic, have been buying physical books, cutting their spines for high-speed scanning, and then destroying them. This decision highlights the growing ethical concerns surrounding data acquisition for AI training, specifically the physical destruction of rare books. By opting for preservation, it sets a precedent for more responsible data gathering practices in the AI industry. While destructive scanning is faster and cheaper because pages can be fed into high-speed industrial scanners, non-destructive methods utilize overhead, flatbed, or V-cradle scanners to digitize pages without dismantling the book.
## BACKGROUND
Large language models require vast amounts of high-quality text data, leading AI companies to digitize physical books. Traditional high-speed digitization often involves cutting the spine of a book to scan individual pages quickly. In contrast, non-destructive scanning technologies, such as V-shaped cradles, support the book's spine and scan pages at a natural angle to preserve the original physical copy.