Research News

Artificial intelligence.

People, machine collaboration key to future of AI


Published May 5, 2016 This content is archived.

Eric Horvitz.
“The future of artificial intelligence is a team between machine intelligence and human intellect. ”
Eric Horvitz, technical fellow and managing director
Microsoft Research Lab

Fusing human perceptions and machines can change the way we make decisions and better address challenges in science, society and daily life, artificial intelligence expert Eric Horvitz told a UB audience on Wednesday.

“The future of artificial intelligence (AI) is a team between machine intelligence and human intellect,” said Horvitz, technical fellow and managing director of the Microsoft Research Lab at Redmond. In his lecture, “Data, Predictions and Decisions in Support of People and Society,” he talked of current research where artificial intelligence was working to help humans understand when AI could be useful in the decision-making process.

The lecture, held in Davis Hall, was sponsored by the Office of the Vice President for Research and Economic Development.

Models that go from data to predictions to decisions also can be used for inference and applied in time-critical situations. Horvitz focused on three areas of Microsoft research that show promise: transportation, health care and people and machines.

Horvitz said systems learning was helpful when working on an app to address the challenges of Seattle traffic. A decade ago, researchers brought together information about road networks, major events and incident reports to determine what traffic would do over time. They added another dataset — one that captured drivers’ thinking and perceptions, as well as “models of surprise” so the app would alert drivers only if they would be “surprised” during their commutes.

In working with health care clients, Horvitz said Microsoft researchers had access to nearly 20 years of data, a rich resource when trying to create models that would help hospitals with the problem of readmissions. Researchers determined from medical charts the discriminating factors in readmission and were able to create what the participating hospitals called the “Microsoft score” that gave the probability of a patient’s readmission. Hospital administrators now can visualize the costs and benefits when simulating patient readmission and make decisions on the level of intervention and post-discharge care.

Another way to assist decision-making is to gather information from objects that already are in play, such as airplanes. Using Microsoft’s Azure Cloud Service, Horvitz said planes flying in different directions can sense the same winds and, with “lots of triangles in the air,” can consider all the variables and receive up-to-date wind and weather information.

Horwitz conducts research on the principles of machine intelligence and leveraging the complementarities of human and machine reasoning. His research and collaborations have led to studies in health care, transportation, human-computer interactions, online services, robotics, operating systems and aerospace.

An award-winning researcher, Horwitz received the Feigenbaum Prize for his contributions to the field of artificial intelligence and is a fellow of the Association for the Advancement of Artificial Intelligence, the Association for Computing Machinery, the American Association for the Advancement of Science and the National Academy of Engineering.

He earned a PhD and MD from Stanford University.

In addition to the lecture, Horvitz took part in an industry/academia roundtable and toured campus during his visit to UB.