UC Davis Law professor Peter Lee discusses the growing use of synthetic data to train AI models and its advantages over real-world data in addressing technical limitations and legal issues like privacy, bias, and copyright infringement. Professor Lee argues that while synthetic data offers promising solutions through unlimited, high-quality training content, it also poses significant risks including model collapse, new biases, and enabling dangerous AI applications, requiring careful regulation and responsible deployment.



