Recent reports indicate that major AI development companies, in an aggressive competition to acquire high-quality text data, have turned to the bulk purchase and destruction of old and rare books. Since web content is increasingly contaminated with low-quality AI productions (AI Slop), old printed books have become a valuable and pure source for training advanced language models.
This process, which involves disassembling book bindings and high-speed scanning of pages to convert them into digital text, has caused deep concern among bibliographers and lovers of written heritage. Although this method is the cheapest and fastest way to digitize massive volumes of information in the shortest possible time, it results in the physical destruction of copies that may never be replaced. Critics warn that these companies are sacrificing a significant portion of human history and material culture in their pursuit of technological progress.