TechCrunch reports that Amazon, a company that began by selling books, is destroying rare texts in order to train AI models. According to the item, such books are considered especially valuable for training large language models because these systems have already been trained on much of the material available online.
Why it matters
Rare texts represent a source of data that is not otherwise accessible online, which the report identifies as the reason for their value to model developers. The described practice of destroying these texts raises questions about how AI training data is sourced.
Who should care
Those following AI training data practices, the handling of rare or physical books, and the companies involved in building large language models may find this development relevant.