In a world where AI is rapidly advancing, a recent court case has shed light on a controversial practice: the destruction of physical books for the sake of training language models. This raises profound questions about the value we place on human creativity, the role of books in our culture, and the ethical boundaries of AI development.
The Case of Anthropic and Project Panama
Anthropic, an AI company, embarked on a mission to improve its large language model, Claude, by employing a method they dubbed "Project Panama." The goal was simple: obtain a vast amount of high-quality language data, preferably pre-2022, to enhance Claude's predictive capabilities. Books, with their intricate narratives and well-crafted language, became the target.
The Choice: Piracy or Destruction?
Anthropic faced a dilemma. They could have sought copyright permission to use existing e-books, but that involved a complex legal process. Instead, they initially opted for piracy, a decision that later led to a $1.5 billion out-of-court settlement with authors. When piracy became too problematic, Anthropic turned to destructive scanning.
The Logistics of Destruction
Destructive scanning, as the name suggests, involves physically destroying books to digitize their content. Anthropic hired logistics experts, sourced books from vendors, and set up a warehouse to house them. The books were then sliced, scanned, and the original copies discarded. It was a well-organized operation, but one that raises ethical concerns.
The Cultural Significance of Books
Books are more than just vessels of information. They hold cultural significance, representing a range of meanings and values. The image of Donald Trump holding a Bible during protests is a powerful example of the symbolic role books can play. Yet, in the US, there are few legal protections for cultural heritage, especially when it comes to books.
The Future of Generative AI and Book Preservation
The court's decision in Bartz v. Anthropic PBC sets a precedent for the use of destructive scanning. If other AI companies follow suit, what does that mean for the preservation of rare and unique printed works? There is a lack of regulation and understanding of textual objects as cultural assets. Should we consider the "wholly human author" as a category for collection preservation?
The Value of Readers and Writers
Judge William Alsup's comments highlight the importance of books and readers. We should be concerned that Anthropic found it easier to destroy physical books than navigate legal processes. This case serves as a reminder of the value we place on human creativity and the potential risks of ceding control to AI models and their proprietors.
A Call to Action
In response to cases like Bartz v. Anthropic PBC, we must refuse to devalue our roles as readers and writers. We must claim ownership of the cultural spaces where our thoughts and words thrive. It's a call to protect and preserve the human element in an increasingly AI-driven world.
