MIT Professor Aleksander Madry strives to build machine-learning models that are more reliable, understandable and robust.
