Abstract: Recent progress in AI and machine learning is stimulating interest in applying AI in high stakes applications such as self-driving cars, surgical robots, and autonomous weapons systems. These applications require high levels of software assurance and resilience, but virtually all AI research has focused on raw performance without paying attention to questions of robustness and resilience.

In this talk, I will survey AI research that aims to create robust systems. I will consider both robustness to “known unknowns” and robustness to “unknown unknowns” — that is, to unmodeled aspects of the environment.

One technology that is relevant to creating robust AI systems is anomaly detection. In the second part of the talk, I will survey the work at Oregon State on anomaly detection. I’ll discuss our recent research on applying anomaly detection to problems of fraud detection and equipment diagnosis and discuss methods for explaining anomaly alarms to an analyst and incorporating analyst feedback into the anomaly detection process.

Bio: Dr. Tom Dietterich (AB Oberlin College 1977; MS University of Illinois 1979; PhD Stanford University 1984) is Professor Emeritus and Director of Intelligent Systems Research in the School of Electrical Engineering and Computer Science at Oregon State University, where he joined the faculty in 1985. Dietterich has devoted his career to machine learning and artificial intelligence. He has authored more than 180 publications and two books. His research is motivated by challenging real world problems ranging from personal information management, to drug design, to sustainability, and most recently to problems in safe and robust artificial intelligence. Dietterich has also devoted many years of service to the research community. He is Past President of the Association for the Advancement of Artificial Intelligence, and he previously served as President of AAAI (2014-16) and as the founding president of the International Machine Learning Society (2001-08). Other major roles include Executive Editor of the journal Machine Learning (1992-98), co-founder of the Journal for Machine Learning Research (2000), and program chair of AAAI 1990 and NIPS 2000. He is currently the moderator for machine learning on arXiv. Dietterich is a Fellow of the ACM, AAAS, and AAAI.