News

U.S. District Judge Vince Chhabria on Monday offered a full-throated denial of one of the authors’ core theories that Meta’s AI system is itself an infringing derivative work made possible only by information extracted from copyrighted material. “This is nonsensical,” he wrote in the order. “There is no way to understand the LLaMA models themselves as a recasting or adaptation of any of the plaintiffs’ books.”

Q: Senate Majority Leader Chuck Schumer (who’s pushing for AI regulations) says making AI models explainable is a priority. How realistic is that?A: It’s a technical feature of these deep-learning, machine-learning systems, that they are opaque. They are black box in nature. But most of the risks we deal with as human beings come from things that are not explainable. As an example, I take a medicine every single day. While I can’t actually predict exactly how it’s going to interact with the cells in my body, we have found ways to make pharmaceuticals safe enough. Think about drugs before we had clinical trials. You could hawk some powder or syrup and it might make you better or it might kill you. But when we have clinical trials and a process in place, we started having the technical means to know enough to start harnessing the value of pharmaceuticals. This is the journey we have to be on now for artificial intelligence. We’re not going to have perfect measures, but I think we can get to the point where we know enough about the safety and effectiveness of these systems to really use them and to get the value that they can offer.