How to join my research group


Note (November 2023): I am currently focusing my research efforts on AGI-related problems such as:

For an introduction to the kinds of problems I'm hoping to help us avoid, see A Taxonomy and Analysis of Societal-Scale Risks from AI.

If you're interested in working with me on probabilistic models, neural SDEs, gradient estimators, or other directions that I've previously worked on, please understand that while I love those directions and hope to get back to them someday, I'm not planning to spend much time on them in the near future, except insofar as they're likely to help address the problems discussed above.

Former Students and Postdocs:

How to apply as a student:

If you'd like to do a Master's or PhD with me, you can apply either through the Computer Science department or the Statistics department, depending on what type of classes you'd rather take. Most of my graduate students are in the CS program.

What are my chances?

Probably not good, unfortunately. There is a shortage of machine learning professors, since so many have gone to industry. I usually take about 1 graduate student per year. That means that pretty much all students who I admit have already done closely related research. Never say never, but it's probably worth applying only if you've already done research related to an area that I work on. I realize that this partly defeats the purpose of graduate school.

What about visiting?

I'm not categorically against it, but only accept visitors under exceptional circumstances.

How to apply as a postdoc:

You should apply through the Vector Institute and put my name as a potential host.

Should I email you directly?

There's nothing wrong with that, however I probably won't reply unless I can see you've put some thought into it. Please explain why you want to work with me specifically, what sort of things you've worked on, and what you'd be interested in learning more about or working on. Even better, explain which approaches you think are currently overhyped or underhyped. If you put the word "shibboleth", written backwards, in your email subject, I'll know that you've read this page.

Other potential advisors with similar interests:

In addition to my collaborators in Toronto, here is a partial list of active professors whose research interests overlap substantially with mine, and whom I know well enough to recommend as potential advisors:


North America: Europe: Asia:
There are great people everywhere, and this list is skewed towards people I have personally worked with. There also many great mentors in industry who I encourage my students to visit and work with.

What if I'm just getting started?

A couple of books that greatly influenced my thinking early on, discuss fundamental themes, are accessible without much background, and free to download are: Good luck!