Meta: Training Llama 3 with Controversial Data

In the case of Artificial Intelligence Copyright in Kadrey v. Meta, the court made public the internal correspondence records of Meta executives and researchers. The records show that Meta, when developing Llama 3, will have as its core goal GPT-4 that goes beyond OpenAI, and that Claude from Anthropic will also be seen as a pole, showing a strong competitive attitude by turning a blind eye to the open source competition, Mistral. In order to achieve its objective, its AI team was “very radical” when it acquired training data, accused of using copyright-protected books to train Llama 3, and discussed internally the use of the LibGen data set to optimize Llama 3 performance, a radical strategy that raised legal issues。