Is it worth to do PhD in machine learning/statistics for work in industry? [closed]

Question

I am 30 years old, have Master's degree in Machine Learning from a CS department. I also have a job, which I have had for like 4 years, first as a software developer, now as a data scientist at the same company.

I have been thinking about PhD for a while now. Is it worth in the field of Machine Learning or statistics (though I think my maths might be too weak for the latter)? My prime reason to do it is that I think that my knowledge in the ML/stats field is not enough, so I would like to dedicate a few years to learning much more even at the expense of low income, and do some paper publishing to meet the quota of "strong publication history". Plus I like to do research, though I like it in industry setting more, since problems are more alive, if you wish.

My particular concern is the work placement after getting my PhD. Many people say that PhD limits the opportunities of a person holding it, but is it true if I already have some industry work experience plus going to apply for ML jobs, unless hype completely disappears in 4 years?

Answer 1

You can answer this question better than almost anyone, because you already know what you need for your work in industry.

I suggest paying close attention to your work environment and thinking about questions such as:

• For what you're currently doing, will having a PhD in machine learning / statistics help?
• For your job, will having a PhD allow you to do something else (preferably something more advanced, more technical, and commands higher pay)?
• Do you have any colleagues with a PhD? If so, are they more effective workers than you because of superior knowledge about ML/statistics? Do they get different responsibilities that you find attractive? Do they get offered those responsibilities (instead of you) because they're more capable than you? Because they know more than you? Because they have PhDs and you don't?
• How much more can you reasonably expect to earn if you have a PhD?

The comparison is easiest if you have colleagues with PhDs, because the most obvious result of you acquiring a PhD is that you become like them in terms of job responsibilities, pay, etc. Even if you don't have colleagues with PhDs however, you can still draw conclusions. For example, if you look at the problem you're currently solving and think, "boy, I can do this so much better if I just understood this algorithm" then that's a sign that a PhD can be helpful. Your current supervisor is someone else whom you can discuss your long-term goals with. Who knows - maybe for the problem you're currently working on, your competitors have people who understand the algorithm and that's why they're gaining market share. In that case your supervisor might actively encourage you to try for it.

tl; dr: use your knowledge of the local industry and job market to figure out what jobs you're likely to be doing if you get a PhD, then decide if you want to do those jobs enough to commit to doing a PhD.

Answer 2

Don't take this as the final word, but I worry that a doctorate, being a very narrow research study, might make you less desirable for most jobs, rather than more. There are companies, of course, that will want researchers focused on a small area, but they are fewer than those with more general needs.

You can, of course, study on your own or take courses to expand your knowledge and skill without an additional degree, but a doctorate is more suited for a life of research, often in academia. Some large companies (IBM, Google, ...) have room for a few true pure researchers, of course. Even there, the research may be quite applied, nowadays. Sun, before they failed, had a very active research department, but one of the reasons for the failure was that they spent a lot of money on that and it didn't come back to them in profits. Not because it was bad research, in fact it was excellent, but it was just research with a long payoff horizon.

On the other hand, there is nothing wrong with learning for its own sake. Some people are just driven to learn about some particular thing.

One thing you can do before you jump into a doctoral program is to do some exploration at your current company and with colleagues you have met from other companies. What do they think about having a researcher? Are there any plans in place to support that? How would your job change if you had a doctorate?