Representations of probability and utility lie at the heart of a two-decade rolling revolution in machine learning and intelligence. A confluence of advances has led to an inflection in our ability to collect, store, and harness large amounts of data for generating insights and guiding decision making in the open world. Beyond study and refinement of principles, fielding real-world systems is critical for testing the sufficiency of algorithms and implications of assumptions—and exploring the human dimension of computational solutions and services.
In this keynote from the Microsoft Research Faculty Summit 2012, Horvitz discusses efforts on learning and inference, highlighting key ideas in the context of projects in transportation, health care, and citizen science. Next, he describes the composition of integrative solutions that draw upon a symphony of skills and that operate over extended periods of time.