In the fall of each year I get a lot of inquiries from people who want to gain admission to one of the graduate programs (master’s or Ph.D.) in Carnegie Mellon University’s School of Computer Science (SCS). Some of these inquiries are just form letters, apparently broadcast to hundreds of faculty members at many different schools; others show some familiarity with my current research. Usually these are accompanied by some sort of resume, and often the students look very good. Sometimes the students request some estimate of their chances for admission or want me to advise them on what additional preparation they need.
We do want to encourage the best students in the world to apply and ultimately to come to CMU. Unfortunately, the number of such inquiries makes it impossible for me to give a thoughtful, individual answer to each of them. So here is a composite answer, addressing most of the questions that students have raised.
1. In our system in SCS, admission to our various graduate programs is handled by an admissions committee in each department, not by individual faculty. Faculty do make known to the committee(s) how many new students we would like to take on in the coming year, the general areas in which we want students, and how we plan to fund any students we do take on. We may write a recommendation letter for student applicants that we actually know from personal contact, but until all admissions materials have been received, that is the extent of our input on individual students. After the materials are all in hand, then we may review the relevant folders and indicate which students we would most like for our own projects.
2. So, unless you personally know a faculty member, it does you no good to contact CMU faculty directly to lobby for your case. It especially does you no good to send a generic form letter saying that you’re interested in something like “AI, theory, robotics, and graphics”. Just apply to the SCS programs of your choice and try to write a good statement of purpose, indicating clearly your interests, skills, and experience.
3. I’m not in a position to comment on your chances for admission to one of our grad programs until we have received the complete application package, including recommendation letters.
4. The admissions committees want to answer two basic questions: (1) Can this person pass our graduate courses without a lot of drama? For that, we look at past grades and test scores, and perhaps take into account any mitigating circumstances. And (2) is there clear evidence that this person can do good research, and not just get good grades in courses? For that we look at publications, previous experience in research projects (at school, as a summer intern, or whatever), and the recommendations. Most of our applicants do well on the first question, so the second one is far more important.
5. If, before applying, you have the opportunity to work on some sort of research project in your university or in industry, it is a good idea to seize that opportunity and to make the most of it. If you can work with a well-known researcher, that makes it easier for us to evaluate any recommendation they might write.
6. Another important factor in admissions, for students whose native language is not English, is whether we are confident that your command of written and spoken English is sufficient to allow you to function in our courses and research projects, and ultimately to write and present good papers. We’ve had bad experiences with students whose English looked marginal on paper (borderline TOEFL scores, etc.), and who quickly failed in our programs after we admitted them. In some cases, we may try to have a phone conversation with the applicant if they look very good, but we’re not sure about the English.
7. If you come from a little-known school or don’t have a very solid CS background, you might consider trying to get into a good but less-selective school to take additional courses and do some research, and perhaps to get a master’s degree. Then apply to CMU for a Ph.D. or a second master’s program (which often can lead to Ph.D. admission later). It’s a longer path, but many have been successful in following it. Another path is to gain some experience in industry before applying.
8. Students who indicate some flexibility and a reasonably broad range of interests will have a better chance of admission that students for whom there is only one plausible advisor. If that advisor doesn’t have funding the year you apply, or prefers some other applicants, then we’re unlikely to admit you, since you would have nobody to work with. So if that flexibility is there, do describe it in your statement of purpose. But don’t stretch the truth too much – it’s not good for you or for us if we admit you and then you can’t find an advisor whose work interests you.
About working for me on the Scone project:
1. If you really are interested in my work, understand that I have not done research in artificial neural networks or statistical machine learning in recent years. My work is now almost exclusively in the area of symbolic knowledge representation and its applications, including knowledge-based natural language understanding. So don’t say you’re passionate about working with me on artificial neural nets. It’s not going to happen.
2. I am most closely associated with the Language Technologies Institute and the Computer Science Department, so it would make the most sense to apply to one of the programs in those departments. However, it is allowed for students in any program to be advised by any faculty member in SCS, so I could in theory advise and support a student in some other department. The choice of department should probably be governed by which set of course requirements makes the most sense for you and your goals.
3. I am currently interested in taking on a few new grad students. I have time and attention for a few additional students, and lots of good project ideas. Admission for the fall of 2013 is now closed. I have no idea whether I will have any funded slots for new students in the fall of 2014. Even if I do get funding for one or two new students, I expect that there will be some strong competition for these positions, both from newly entering students and from some students already here. So unless you have a pretty solid background in AI and knowledge representation and can implement your ideas, it is unlikely that you would be invited to join our group. If you have some external source of funding – a fellowship, government funding from your country, or personal/family money – the odds for you are much better, since you would not be competing with many others for an existing (or non-existent) funding slot.
4. If you want to find out more about my work and overall approach to AI, look at the Scone Project home page (Google “Fahlman Scone”) and my Knowledge Nuggets blog (Google “Fahlman Knowledge Nuggets”).
So, good luck in finding the best program for you. And if it makes sense, don’t be afraid to apply to one or more of our graduate programs at Carnegie Mellon to see what happens.