Michael I. Jordan
Last reviewed
Jun 8, 2026
Sources
15 citations
Review status
Source-backed
Revision
v1 · 2,056 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Jun 8, 2026
Sources
15 citations
Review status
Source-backed
Revision
v1 · 2,056 words
Add missing citations, update stale details, or suggest a clearer explanation.
Michael Irwin Jordan (born February 25, 1956) is an American scientist who ranks among the most influential and most cited researchers in machine learning, statistics, and artificial intelligence. For more than two decades he was the Pehong Chen Distinguished Professor in the Department of Electrical Engineering and Computer Sciences and the Department of Statistics at the University of California, Berkeley, where he is now an emeritus professor, and since 2023 he has also held an appointment as a directeur de recherche (research director) at Inria and the Ecole Normale Superieure in Paris. [1][2] His Google Scholar profile lists more than 367,000 citations and an h-index above 220, figures that place him among the most cited scientists in any field. [3] In 2016 a Semantic Scholar analysis of research influence ranked him the single most influential computer scientist then active, prompting the journal Science to ask, only half in jest, who was "the Michael Jordan of computer science." [4]
Jordan is best known for helping to recast machine learning as a rigorous inferential and engineering discipline grounded in probability theory, probabilistic graphical models, and optimization. He was a central figure in popularizing variational inference and the expectation-maximization algorithm, and he co-authored the topic model known as Latent Dirichlet Allocation, one of the most cited papers in the field. [1][5] He is equally known for the breadth of his academic lineage: his former students and postdoctoral researchers include several of the people who went on to lead modern AI. [1]
Jordan was born on February 25, 1956, in Aberdeen, Maryland. [1] He took an unusually broad path into the field. He earned a Bachelor of Science in psychology from Louisiana State University in 1978, graduating magna cum laude, then a Master of Science in mathematics from Arizona State University in 1980. [1] He completed his doctorate in 1985 at the University of California, San Diego, where he worked in cognitive science under the psychologists David Rumelhart and Donald Norman, two founders of the connectionist, or parallel distributed processing, school of thought. [1]
His doctoral and early research concerned how the brain represents serial order in behavior. In the mid-1980s he introduced an early form of recurrent neural network, later called the "Jordan network," in which a network's output is fed back as context for the next time step. [1] This mix of psychology, mathematics, and cognitive modeling shaped a career spent translating between the statistical and the computational views of learning.
Jordan joined the faculty of the Massachusetts Institute of Technology in 1988, where he was a professor in the Department of Brain and Cognitive Sciences until 1998. [1] In 1998 he moved to UC Berkeley, taking a joint appointment between the Department of Electrical Engineering and Computer Sciences and the Department of Statistics, and eventually holding the Pehong Chen Distinguished Professorship. [1][2] At Berkeley he became one of the intellectual anchors of the campus's machine learning and data science community, affiliated with the Berkeley AI Research lab, the Sky Computing Lab, and the CLIMB center for the theoretical foundations of learning, inference, and microeconomics. [2]
He formally became an emeritus professor at Berkeley in 2024, after roughly twenty-six years on its faculty, while remaining research active. [2] In 2023 he took up a position as a directeur de recherche at Inria, the French national institute for digital science, working in Paris in association with the Ecole Normale Superieure. [2]
| Period | Role | Institution |
|---|---|---|
| 1978 | B.S., psychology (magna cum laude) | Louisiana State University |
| 1980 | M.S., mathematics | Arizona State University |
| 1985 | Ph.D., cognitive science | UC San Diego |
| 1988 to 1998 | Professor, Brain and Cognitive Sciences | MIT |
| 1998 to 2024 | Pehong Chen Distinguished Professor (EECS and Statistics) | UC Berkeley |
| 2023 to present | Directeur de recherche | Inria and ENS, Paris |
| 2024 to present | Pehong Chen Distinguished Professor Emeritus | UC Berkeley |
Jordan's research has repeatedly bridged communities that had developed separately. In the 1990s and 2000s he was a leading proponent of probabilistic graphical models, a framework that uses graphs to encode the conditional independence structure of large collections of random variables. He helped connect the artificial intelligence tradition of Bayesian networks with the statistics tradition of exponential families and the physics tradition of mean-field methods. [1]
A signature theme is approximate inference. With his students and collaborators he popularized variational inference, in which an intractable probability distribution is approximated by a simpler one chosen to minimize a well-defined discrepancy, turning inference into an optimization problem. His 1999 tutorial "An Introduction to Variational Methods for Graphical Models," written with Zoubin Ghahramani, Tommi Jaakkola, and Lawrence Saul, became a standard reference, as did the later monograph "Graphical Models, Exponential Families, and Variational Inference" (2008) with Martin Wainwright. [6][7] He also did influential work on the expectation-maximization algorithm and on combining it with neural-network-style "mixtures of experts." [1]
Two of his most cited papers came from the early 2000s. With his doctoral students David Blei and Andrew Ng, he introduced Latent Dirichlet Allocation in 2003, a generative probabilistic model that discovers latent "topics" in collections of documents. It became one of the foundational tools of text analysis and has been cited tens of thousands of times. [5] With Ng and Yair Weiss he wrote "On Spectral Clustering: Analysis and an Algorithm" (2002), which helped make spectral methods a standard clustering technique. [8]
Across this body of work runs a consistent argument: that machine learning should be built on the union of computer science and statistics, with explicit attention to uncertainty, decision making, and the trade-off between statistical accuracy and computational cost. In more recent years Jordan has extended this program toward economics, studying markets, incentives, and "learning-aware" mechanism design, on the view that large learning systems are really collections of interacting economic agents and should be analyzed as such. [2][9]
Beyond his own papers, Jordan is famous for the scientists he trained. The reputation that he "advised a generation of leaders" is accurate, though the relationships are often mischaracterized: several of the best known figures were postdoctoral researchers rather than doctoral students. [1] Yoshua Bengio, a 2018 Turing Award laureate, was a postdoc with Jordan, as was Martin Wainwright. By contrast, Andrew Ng, David Blei, and Zoubin Ghahramani were among his doctoral students. [1]
| Name | Relationship | Later known for |
|---|---|---|
| Andrew Ng | PhD student (Berkeley) | Google Brain, Coursera, Stanford |
| David Blei | PhD student (Berkeley) | Topic models, Columbia University |
| Zoubin Ghahramani | PhD student (MIT) | Cambridge, VP of Research at Google DeepMind |
| Eric Xing | PhD student (Berkeley) | Carnegie Mellon, president of MBZUAI |
| Tamara Broderick | PhD student (Berkeley) | Bayesian nonparametrics, MIT |
| Yoshua Bengio | Postdoc (MIT) | Deep learning, 2018 Turing Award |
| Martin Wainwright | Postdoc (Berkeley) | High-dimensional statistics, MIT |
These and other advisees populate the faculties of Berkeley, Stanford, MIT, Carnegie Mellon, Columbia, and Cambridge, as well as the research labs of Google, and many in turn trained the next cohort, giving Jordan an outsized influence on the field that extends well beyond his own publications. [1]
Jordan has become one of the most prominent skeptics of "artificial intelligence" as a description of what the field actually does, even as he remains one of its central contributors. In a widely read 2019 essay in the Harvard Data Science Review, "Artificial Intelligence: The Revolution Hasn't Happened Yet" (first circulated in 2018), he argued that public discussion conflates several distinct goals, and that the genuine, world-changing advances had come not from human-imitative AI but from large-scale systems for recommendation, search, fraud detection, logistics, and similar tasks. [9] What was needed, he suggested, was a new engineering discipline for systems that learn from and act on data, comparable to chemical or civil engineering and drawing on statistics, optimization, control, and economics as much as on computer science. [9]
He has repeatedly cautioned against hype and imprecise language. In a 2021 IEEE Spectrum interview headlined "Stop Calling Everything AI," he noted that the backpropagation algorithm at the heart of modern deep learning traces back to control theory and gradient methods of the 1950s and 1960s, and he warned that overpromising on human-like intelligence risks both a public backlash and the neglect of more tractable and valuable problems. [10] Consistent with these commitments to the health of the field, in 2001 he resigned from the editorial board of the journal Machine Learning to support the newly founded, open-access Journal of Machine Learning Research. [1]
Jordan is one of relatively few researchers elected to both the National Academy of Sciences (2010) and the National Academy of Engineering, and he is a member of the American Academy of Arts and Sciences and a Foreign Member of the Royal Society (2021). [1][11] He is a Fellow of the ACM, the IEEE, AAAI, the Society for Industrial and Applied Mathematics, the Institute of Mathematical Statistics, and the American Statistical Association. [1]
His major prizes span computer science, statistics, and mathematics, an unusual breadth that mirrors his cross-disciplinary work.
| Year | Award |
|---|---|
| 2015 | David E. Rumelhart Prize |
| 2016 | IJCAI Award for Research Excellence |
| 2020 | IEEE John von Neumann Medal |
| 2021 | AMS Ulf Grenander Prize [14] |
| 2021 | Mitchell Prize (with co-authors) [15] |
| 2022 | World Laureates Association Prize (inaugural) |
| 2025 | BBVA Foundation Frontiers of Knowledge Award |
The 2015 Rumelhart Prize formed a notable bookend, honoring Jordan with the award named for his own doctoral advisor. [1] The 2021 Ulf Grenander Prize from the American Mathematical Society cited his "foundational contributions to machine learning, especially unsupervised learning, probabilistic computation, and core theory for balancing statistical fidelity with computation." [14] In 2022 he became the inaugural laureate of the World Laureates Association Prize in Computer Science or Mathematics, which carries a monetary award of 10 million RMB (about 1.4 million US dollars), cited "for fundamental contributions to the foundations of machine learning and its application." [12] In 2025 he shared the BBVA Foundation Frontiers of Knowledge Award in Information and Communication Technologies, endowed with 400,000 euros, with the biometrics researcher Anil Jain. [13] As of 2026 he continues to hold his emeritus chair at Berkeley and his research directorship in Paris, and he remains an active researcher and a frequent public voice on the foundations and direction of machine learning. [2][13]