Thomas G. Dietterich is an American computer scientist and emeritus professor of computer science at Oregon State University. He is recognized as one of the pioneers of the field of Machine learning, having contributed foundational algorithms and frameworks that shaped modern artificial intelligence research. His work spans multi-class classification, hierarchical reinforcement learning, and the integration of statistical methods into probabilistic models, with applications ranging from ecology to computer security.
Dietterich also played a key role in the academic infrastructure of machine learning, serving as executive editor of the journal Machine Learning from 1992 to 1998 and helping to co-found the Journal of Machine Learning Research. He has been a prominent voice in public discussions about the risks and benefits of Artificial intelligence, offering an academic perspective to media outlets such as National Public Radio, Business Insider, Microsoft Research, CNET, and The Wall Street Journal.
Early life and education
Thomas Dietterich was born in South Weymouth, Massachusetts, in 1954. His family relocated to New Jersey and later to Illinois, where he graduated from Naperville Central High School. He then attended Oberlin College, graduating in 1977 with a degree in mathematics, focusing on probability and statistics. During the summer of 1977, he worked as an assistant to the Director of Planning and Research at Oberlin College.
Dietterich spent the next two years at the University of Illinois, Urbana-Champaign, before beginning doctoral studies in computer science at Stanford University. He received his Ph.D. in 1984, after working as a research assistant in the Heuristic Programming Project from 1979 to 1984. In the summer of 1979, he also held a position as a member of technical staff at Bell Telephone Laboratories in Naperville, Illinois, where he worked on computer-to-computer file transfer and micro-code distribution to remote switching systems.
Academic career at Oregon State University
In 1984, Dietterich moved to Corvallis, Oregon, and joined Oregon State University as an assistant professor of computer science. He was promoted to associate professor in 1988 and to full professor in 1995. In 2013, he was named Distinguished Professor, a title he held until his retirement in 2016. He also served as Director of Intelligent Systems Research in the School of Electrical Engineering and Computer Science from 2005 onward.
During his career, Dietterich took several sabbatical and industry positions. From 1991 to 1993, he was a senior scientist at Arris Pharmaceutical Corporation in South San Francisco. He served as chief scientist at MyStrands, Inc. (2004–2005) and at Smart Desktop, Inc. (2006–2008). In 1998–1999, he was a visiting senior scientist at the Institute for the Investigation of Artificial Intelligence in Barcelona, Spain. Since 2011, he has been chief scientist at BigML, a machine learning company based in Corvallis.
Research contributions
Dietterich's research has produced several influential methods in machine learning. He invented error-correcting output coding, a technique for improving multi-class classification by decomposing the problem into multiple binary classifiers and using error-correcting codes to combine their outputs. He also formalized the multiple-instance problem, a learning paradigm where labels are associated with bags of instances rather than individual instances, which has applications in drug design and image analysis.
In reinforcement learning, Dietterich developed the MAXQ framework for hierarchical reinforcement learning, which decomposes a complex task into subtasks and enables more efficient learning and planning. He also contributed methods for integrating non-parametric regression trees into probabilistic graphical models, bridging the gap between statistical machine learning and structured prediction.
Dietterich has expressed interest in all aspects of machine learning, with three major strands: building integrated intelligent systems, enabling human-computer collaboration, and applying machine learning to ecological sciences and ecosystem management. His work in computational sustainability includes projects on wildfire management, invasive vegetation, and bird migration modeling. For example, he collaborated with the Cornell Lab of Ornithology to analyze data from eBird, a citizen science platform that collected over 3.1 million bird observations by March 2012, using machine learning to uncover migration patterns.
Professional service and leadership
Dietterich has been deeply involved in the organization of the machine learning and AI communities. He served as technical program co-chair of the National Conference on Artificial Intelligence (AAAI-90), technical program chair of the Neural Information Processing Systems conference (NIPS-2000), and general chair of NIPS-2001. He was the founding president of the International Machine Learning Society and has remained a member of its board since its founding. He also serves on the steering committee of the Asian Conference on Machine Learning.
For many years, Dietterich was editor of the MIT Press series on Adaptive Computation and Machine Learning, and he co-edited the Morgan Claypool Synthesis Series on Artificial Intelligence and Machine Learning. He has also been a member of the Association for the Advancement of Artificial Intelligence (AAAI), serving as its president from 2014 to 2016.
Perspectives on AI risks
Dietterich has offered a measured perspective on the dangers of Artificial intelligence, arguing that the most realistic risks are basic mistakes, breakdowns, and cyberattacks, rather than machines becoming super powerful or destroying humanity. He considers scenarios of self-aware machines exterminating humans to be more science fiction than scientific fact. However, he acknowledges that as computer systems are given increasingly dangerous tasks and asked to learn from experience, they may make errors. He notes that much of the work in the AI safety community focuses on accidents and design flaws, aligning with his view that technical reliability is the primary concern.
Legacy and influence
Dietterich's contributions have influenced both theoretical and applied machine learning. His error-correcting output coding and MAXQ framework are widely cited and used in various domains. His advocacy for computational sustainability has inspired researchers to apply machine learning to environmental challenges. Through his editorial roles and conference leadership, he helped establish the infrastructure for the modern machine learning community, which has grown into a central field of Deep learning and Generative AI research.
Dietterich's career reflects a commitment to rigorous science and practical impact. After retiring from Oregon State University in 2016, he continued to engage with the field through advisory roles and public commentary. His work remains a reference point for researchers addressing complex classification and reinforcement learning problems, and his perspective on AI risks continues to inform discussions about the safe deployment of intelligent systems.