I was admitted to several Phd position and I am considering most interesting two labs. Their fields are similar and both interesting to me. The problem is, one lab has more general topic, but the other has more specific topic.

The 'A' lab is focusing on brain-inspired AI. It aims to develop deep learning model based on neuroscience. There is not much rooms for mathematics seeing the previous papers. The 'B' lab is more concentrated on machine learning itself and uses extensive mathematics.

After I get Phd in CS, I will probably go to industry or perhaps to academia and I should be able to apply my knowledge to new domains which is required by company. I think mathematics is important for this flexibility.

I am worrying that if I study brain-driven AI during Phd, I might lack flexibility since the topic is specific and it is not using math a lot. Yes, I know that as a Phd, I should study mathematics by myself required for machine learning but it is also true that student whose topic is more math oriented ML will relatively have better knowledge at the end of Phd.

Some give me advice that I should go to a lab having more broad field of study. And it is better to go to more specific fields as a post-doctor. Considering the fact that I should always study new things by myself after Phd, I think it would be better to be trained with extensive mathematics which is core baseline of ML.

What do you think?

3 Answers 3


There are two broad criteria for a PhD: it must advance the state of the art (novelty); and it must be your own work (independence). The novelty criterion tends to suggest that all research PhDs are 'narrow'. The more interesting question is whether the field is 'crowded' (too many researchers chasing few new questions).

Congratulations on your admission to multiple PhD positions. I'd suggest that you consider the following, with the understanding that you make your own decisions on this:

  • Is the topic of personal interest to you? This can be for the academic content or for job prospects or skills likely to be gained, etc.

  • Does the supervisor have a good track record of having doctoral candidates that graduate within a reasonable time?

  • Is there a reasonable expectation that you'd be able to settle on a research topic that is solvable and will satisfy the novelty aspect when you write up?

These questions help you deal with motivation, completion and realistic expectations.


That's a very difficult question to answer, and my suggestions below are speculative, but based on experience and observations.

Specific project: On the other hand, a more specific project will be easier to address, you will spend less time wandering around, and you will gain specific skills that you can apply and reapply. These skills are more likely what industry will want.

General project: I think any project that offers a more general topic will give you greater flexibility to make your mark on the world, if you have the capabilities and the luck to come up with the right ideas at the right time. Making your mark is beneficial for subsequent academic positions, and essential if you want to be a big-shot celebrity.

The risk with general topics is that you have too much freedom and, without case studies or something to draw inspiration from, you may be forced to guess a lot of the time. That is fine, science is guessing, but given that you will be a PhD student you will likely lack the intuition to make your guesses worthwhile.

Short-term advice: a specific topic may be an easier path to industry, a general topic may be easier for academia, but you should make sure that the general topic is not too general.

Long-term advice: Working on a general and highly-speculative topic may lead to you becoming an expert in a new area of research, and hence demand from industry, but that would likely be many years after you PhD.


If your goal is to work in industry after the fact, I would look for a specific problem that is not well served by the prior art. Doing general research on AI has some academic sex appeal but I don't think generic AI research will translate well to a position in industry. I would look for professors in other fields (engineering, science, and math) who work on computational algorithms for industrial applications and solicit them for positions.

I did my undergrad in Software Engineering and Management and I found that I had a lot to offer a mechanical engineering PhD who works in reliability of machinery.

Not the answer you're looking for? Browse other questions tagged .