AI companies destroy physical books for data training, says Lawrence Liang

Lawrence Liang, writing for National Herald India, draws a stark parallel between the Nazi book burnings and the current practice of AI companies like Anthropic destroying physical books to train large language models. He describes how industrial blades guillotine book spines, and pages are scanned before the remains are shredded. Unlike past purges driven by political ideology, this destruction is done in the name of knowledge production, treating books as mere containers of raw data.

Liang notes the irony that, in early print history, binders destroyed old manuscripts for their parchment, discarding the text. He argues that replacing human readers with memory machines reverses the hope in Bradbury's Fahrenheit 451, where people preserved books by memory. While digitisation itself is not bad, Liang questions the methods of multibillion-dollar companies that choose destructive scanning for speed and cost. He also raises the uncomfortable question of ordinary complicity in the age of AI.