This podcast episode explores the concept of predicate variables in machine learning and how they are used to decrease the admissible set of functions. It is an essential tool in understanding the essence of dog, among other things.