4

I just started my PhD which is in Artificial Intelligence (AI).

My supervisor is specialized in Electrical Engineering with a hint of AI.

He knows that it is not his area of expertise and suggested if I can find a mentor who is more specialized.

I contacted many professors from other universities but no one accepted to become my mentor.

I'm not sure what to do now, especially because it is not easy to find another university.

Do you have any advice or suggestion for this situation?

  • 10
    Why did you choose a supervisor that is not an expert in AI? I suspect no one will be willing to supervise unless there is a formal arrangement or that someone is an existing collaborator of your supervisor. Your best option is focus on an EE project. – Prof. Santa Claus Dec 19 '18 at 0:40
  • 8
    AI is an innately inter-disciplinary topic. When AI is applied to a real problem, it is just as important (if not more important) to have a domain expert who know the problem inside out as it is to an AI expert. As long as the project isn't deriving fundamentally new methods in AI, but rather applying AI to some domain, then a primary supervisor in the problem domain and a secondary in ML is perfectly suitable. However, the primary supervisor should be the one finding the secondary supervisor. – Ian Sudbery Dec 19 '18 at 12:36
  • @Prof.SantaClaus Coz this is the university that accepted my PhD application, I applied for other Unis but with no much luck, this Dr. accepted my proposal and he told me it will be hard but will be able to do it. now I am listening to the comments here, I am really scared now. – asmgx Dec 19 '18 at 21:23
  • @IanSudbery My PhD is in ML, and my supervisor is not expert in ML :( – asmgx Dec 19 '18 at 21:24
  • I bet you are in Australia, where according to my observation, this is very very very common. I was in this situation myself, along with a bunch of my phd colleagues. What I did: contacted somebody in Europe who kindly agreed to unofficially supervise me and steer me towards finishing. You need to have a good profile for anyone to be willing to invest the time in you though. – Pioneer83 Dec 24 '18 at 6:55
1

I would ask for clarification on this line:

He knows that it is not his area of expertise and suggested if I can find a mentor who is more specialized.

What does he mean by this? Is he willing/eager to supervise you, or not?

  • If not, then I agree with the other answers here, find a different supervisor at all costs
  • But if he is willing to supervise you (and he said this just to warn you that he is not an AI expert), then you could try to make this work. You'll have to agree on a schedule.

    • I recommend that you spend the first year to teach yourself about AI (maybe check out pyimagesearch.com -- the paid materials are excellent, though overpriced...I assume you know how to code already, if not, that will be a problem), get your supervisor to send you to a conference like CVPR or NIPS, and take any courses your university offers. You should cover all the basics like CNNs, LSTMs, and GANs as well as non-DL techniques (e.g., random forest, hidden markov models, etc.). Not everyone has what it takes to keep motivated and teach themselves a new subject for a whole year; hopefully you do (if not, consider switching universities)
    • From there, your EE supervisor should be able to help you identify a research topic of mutual interest -- for example, running these under SWaP conditions, or proving information theoretic theorems, or whatever else EEs care about, I don't know.
2

You face a few options. Suppose that you insist that you must do your PhD in AI.

  • You should do whatever it takes to find a new advisor or an appropriate co-advisor. This includes talking to other faculty, the department chair, and even the dean of the college and graduate school. When "whatever it takes" fails to be a suitable choice, you must take a different approach.

Suppose that you decide or determine that doing a PhD in AI is not as important as doing a PhD at the university where you currently reside.

  • You should do whatever it takes to find an advisor + project that is of suitable interest to you for your work. This especially includes talking to other faculty as well as reading in other research areas. When "whatever it takes" fails to be a suitable choice, you must take a different approach.

Alternatively, suppose that you decide or determine that you will do the PhD in AI with only administrative support and no research support from a faculty advisor.

  • You should do whatever it takes to complete the appropriate research on your own. This includes reading journals and other dissertations in the field and communicating directly with other researchers in the field. This can/should include bringing faculty on to your committee who have the appropriate background in your chosen topic.

Suppose that you decide or determine that you must do your PhD in a certain topic (AI or something else) but for whatever reason cannot do it where you currently are.

  • You should do whatever it takes to find a new university + advisor + project that is of suitable interest to you. This includes researching where your topic is currently being done at other universities and contacting the faculty at that university directly to ask whether they have a potential opening for you to join the group. When "whatever it takes" fails to be a suitable choice, you must take a different approach.

At this point, you may likely realize that you will not be doing a PhD.

2

It really depends on you. If you have good background, have done Masters and knows how to to do research, done some AI experiments, determined and motivated, yes you can do Phd. with your described circumstances. You can design an experiment and submit papers and from there you build your knowledge.

Many students do Phd. in almost totally independent way. Like anything in life, there is an existing ideal image drawn about what/how things should be done. But you find yourself in another circumstances that this ideal image can not be applied.

My rule is: "Do what you can, with what you have, where you are".

You can either:

1) Change the topic/area to something that your supervisor can provide expertise.

2) Continue in what you want, validate your results through conferences.

Start from what you know and keep going.

1

Finding a co-supervisor is a responsibility of your primary supervisor. You should be guided on how to find a qualified mentor. Have you tried your own university? Can you ask help from your EE supervisor? What about asking your school to help (maybe someone else in your school has the connection)?

Sounds like you've just started the degree. If your supervisor and your school are unable to find a mentor for you, please consider suspend your PhD. No point for you doing an AI PhD unless you have someone mentoring you.

  • My supervisor asked other Professors from the same uni to join but they were all busy. – asmgx Dec 19 '18 at 1:00
  • @asmgx are you sure you want to do this phd? – SmallChess Dec 19 '18 at 1:16
  • now that you are making it sound like a big trouble, I started to have doubts that I can continue. but I really really want to do the PhD – asmgx Dec 19 '18 at 1:34
  • @asmgx you can still do it, just much harder when you have no AI supervisor – SmallChess Dec 19 '18 at 1:53
  • 1
    Finding a co-supervisor is a shared responsibility. – Jeffrey J Weimer Dec 19 '18 at 2:20
1

I would also consider if you can just swing the whole thing on your own. You should realize that many professors (even in areas where they share expertise) append their names routinely to papers where all the ideas, work, and even writing came from the grad student.

At the end of the day, you sort of sink or swim on your own. Now, it would be nice to have an expert in your field and some apprenticeship. But academia is drastically far from an apprenticeship model.

If you think you can handle it, just do it all on your own. Might even be easier since the old ma...person won't try to interfere with you. Just add his name to your papers and plunge ahead as a stealth PI!

Honestly even when the fellow is in your field, you may find more collegial interactions with other professors, older students, postdocs, people from adjacent fields and conferences, etc.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.