Court documents connected to the Anthropic case reveal that an internal project to digitize old books was known as “Project Panama.”
According to the documents, the company’s stated aim was to “destructively scan every book in the world” for use in digital libraries and artificial intelligence training.
Destructive book scanning involves sending physical books to specialized facilities, where they are digitized on an industrial scale.
The process typically involves removing each book’s spine so its pages can be scanned rapidly before the remaining material is recycled.
“There are many mysteries surrounding the Panama Plan,” says Mr. Manley of Barter Books.
“While the name is new to me, the reality of this project is not. It has been widely discussed on bookstore forums.”
Mr. Manley says he is unaware of whether his books are being purchased by AI companies for Project Panama or similar book-digitization efforts.
However, he says it can be difficult to explain sales that appear random and have “no rhyme or reason.”
The books reportedly range from obscure Latin texts and rare academic works to classic cowboy novels.
Experts say the wide range of subjects could have important implications for AI development. Rare and unusual books may provide new training data to improve large language models, the technology that powers generative AI tools such as chatbots.
Professor Emily Hudson, an intellectual property expert at the University of Oxford, says UK copyright law differs significantly from US copyright law.
“The starting point in the UK is that all copying activities, including creating training libraries and delivering training, require permission from the copyright holder,” she says.
Source: www.bbc.co.uk