Inside Meta’s race to beat OpenAI: “We need to learn how to build frontier and win this race”

Meta's internal communications reveal plans to use copyrighted data, including from book piracy site LibGen, to train AI models like Llama. The company aimed to match OpenAI's GPT-4 capabilities while implementing measures to conceal data sources and mitigate potential legal and regulatory risks. This approach highlights the competitive race among AI companies to acquire training data amid growing concerns about copyright infringement and data scarcity.