Rethinking Deep Learning

Paul and Keith discuss the need to refound deep learning for better reasoning and theorem proving. They explore the challenges of inscrutability in large-scale ML systems and the potential for a more structured approach akin to classical programming.