Michael I. Jordan

Michael I. Jordan (Aberdeen, Maryland, United States, 1956) holds a master’s degree in Mathematics (Statistics) from Arizona State University (1980) and a PhD in Cognitive Science from the University of California, San Diego (1985). His career has largely been spent at the Massachusetts Institute of Technology (1988-1998), where he was a professor in the Department of Brain and Cognitive Sciences, and, latterly, at the University of California, Berkeley, where he has been Pehong Chen Distinguished Professor in the Department of Electrical Engineering and Computer Science and in the Department of Statistics and is currently Emeritus Professor. He is also a Senior Researcher at the National Institute for Research in Digital Science and Technology (INRIA) in Paris (France), where he heads the Markets and Machine Learning chair. Author of over 230 publications in scientific journals, he has served as president of the International Society for Bayesian Analysis and on the editorial boards of journals including Statistics and Computing and Stochastic Analysis and Applications.

Michael I. Jordan has developed mathematical and computational techniques that underpin a wealth of AI applications. From restaurant recommendation systems to generative language models like ChatGPT or decision support tools in the business domain, Jordan’s advances have been instrumental in enabling their real-life application. In addition to laying the theoretical foundations for such uses, the awardee has brought several of them to market in partnership with companies.

Jordan began his research career looking at the models used to establish probabilistic relations between different variables, which are a key component of text and image analysis and recommendation systems. Back in the 1990s, he was also at the forefront of the development of so-called variational inference models, able to approximate the solution to a mathematical problem that is not solvable with available computational resources, by reducing it to an optimization problem. This technique is a core component of machine learning, particularly deep learning applications like the generative AI of ChatGPT and other language models.

In the 2000s, Jordan turned his attention to multiplying the possibilities of machine learning by running programs on hundreds or thousands of computers instead of just one. The algorithms he devised to enable such large-scale distributed computing led to the setup of the company Anyscale, whose Ray platform is the basis for ChatGPT, numerous e-commerce firms and many other deep learning applications.

Among the awardee’s more recent interests has been the application of machine learning to economics. In contexts where multiple actors are entrusting decisions to the same system, recommender systems must be able to adapt to avoid congestion. For instance, a GPS app being used in a town with hundreds of thousands of inhabitants could recommend the same route to the airport to a thousand users at once, causing traffic hold-ups. Jordan is working to develop machine learning systems that overcome this problem, reflecting people’s preferences while allowing them to collaborate within the same system (for instance, choosing alternative routes so as to generate lighter traffic on each). “What we aim for is to have people and decision-making systems working together and finding solutions that are valuable or appropriate for everybody. The kind of thing that an economist would think about,” he explained in an interview. “It’s less about collecting vast stores of knowledge and knowing everything about the world and telling us the answer. More about making us connect better so we can get more out of each other and collaborate more effectively. I want to empower humans, not have the AI tell humans what to do.”