Meta admits using pirated books to train AI, but won't pay for it
Meta has acknowledged using the Books3 dataset—a collection of over 195,000 books—to train its Llama 1 and Llama 2 large language models (LLMs). This admission came in response to a lawsuit filed by authors, including Sarah Silverman and Richard Kadrey, who allege that Meta utilized their copyrighted works without permission. Meta contends that its use of these materials falls under “fair use,” arguing that it doesn’t require consent, credit, or compensation to the authors. The company also disputes the validity of the lawsuit as a class action and denies any obligation to provide monetary relief to the plaintiffs. (TechSpot)
