12

If I have the option of choosing between a (non-research) internship at Google and a REU program at a non-prestigious university, which one should I select? After graduation, I intend to pursue a PhD in Computer Science, hopefully in machine learning.

Which of the two offers should I accept to improve graduate admissions? What if the intern position is directly in what I want to work in, i.e. machine learning?

I know research is valued more for graduate admissions, but a machine learning internship at google seems like a valuable asset.

  • 2
    While it would likely help, it really depends on what you get to do at Google. – xuq01 Jan 28 at 5:08
  • 1
    What matters is the quality of the recommendation letter you can get as a result. – Thomas Jan 28 at 5:55
  • 8
    In what sense is a non-research internship at Google “prestigious”? – JeffE Jan 28 at 19:34
  • 1
    @JeffE in an industry sense (which would beg the question - what do you want this for?) – JGreenwell Jan 28 at 21:17
  • @xuq01 : this reminds me of a certain minister of social services in a certain country, who used that he worked before at a very prestigious international charity organization, implying he was a manager there... and later it turned out he worked there as a driver. – vsz Jan 29 at 11:52
35

Take the internship. It's not so much about whether it helps for graduate admissions or not, but rather how it is going to broaden your horizons. If you do the REU program, when you start the PhD, you're going to be doing similar things. On the other hand taking the internship gives you firsthand experience you cannot otherwise get:

  • What working in the real world is like
  • What skills are actively useful in industry
  • What skills you currently lack, but can learn in a PhD
  • What kind of job you can do if you get a PhD, but can't do now

(This neglects the financial aspect entirely, plus the fact that you're still going to be learning things during the internship which you can write into your application.) These things are likely to change your view of the PhD entirely. In the extreme case, you might find you don't want to do a PhD anymore. Less extremely, you'll have a better idea why you're studying, what you want to learn, and how that is going to help you.

tl; dr: do the different thing, because it'll act as a great enabler even if it's not immediately helpful.

14

I was exactly in your position a year before finishing my master’s. I was in touch with Google regarding an internship (I've passed their interviews and could've accepted as soon as we matched me with a topic). The timing for the Google internship didn't fit brilliantly with my university schedule (we had the longest semester in Europe back in the time), so I went to talk to my advisor at the time to ask him whether he thinks I should take the internship despite it colliding with three weeks of my classes.

I was at the time seriously considering following up my master’s with a PhD programme, and he knew that when I came for advice. His response was that if that's my plan, I would benefit much more from a research internship – and then he proceeded to arrange one for me (and one more the following year; he really was brilliant and dedicated advisor).

I agree that through an industry internship you will acquire skills Allure mentions in his answer, and I do admit I lack some of them even now. On the other hand, taking the research internships got me into a PhD programme. Some concrete benefits I got from there (somewhat location dependant):

  • I got into my first proper research group. While my master’s advisor was golden, research is not too strong in my country of origin.
  • It got me a non-generic, research-oriented letter of recommendation
  • Inquiring about PhD positions through e-mail has a very low response rate; even if you don't ask about publicly available information and only ask about advertised positions through proper channels.
  • Politely knocking on somebody's door and saying "Hi, I'm doing an internship in the group down the hall and saw you advertising a PhD position. I was wondering if you have five minutes one of these days to talk about it a bit" had a much higher response rate.
  • I got coached for the admissions process by my advisers-to-be, as they decided to support me as a candidate following that five-minute conversation mentioned in the previous point.

I do have a feeling the benefits were quite tangible as I come from a fairly small country, my master’s university was not particularly famous and there is not much research support. Thus, being at a research institute and in direct contact with several research groups gave me connections and opened some doors for me which would have been much more difficult to achieve otherwise.

All that said; taking a machine-learning internship at Google will surely not harm your profile or your CV. Just remember – it is a software engineering internship and therefore will give you those skills; if you want to focus on your research skills it is not a best choice. If you want to experience a bit of everything before committing (to a PhD, say), it just might be.

2

Acceptance on any program is, or should be, based on your attributes, capacities and skills. Ie it is not where you did the internship, but what you learnt from it...

1

The answer depends on some factors not mentioned in the question. Some points to consider:

1) I'd say, as a rule PhD programs are not overcrowded, so I am not sure whether "improving the chance of admission" is an issue at all for a reasonably skillful student. (Unless you are aiming at some really high-profile institution and you know for sure that there is a serious competition even at PhD level).

2) However, if you are planning to obtain financial support for your PhD studies, it's a completely different story. Competition for scholarships can be severe, but again, if you know that a certain department offers N scholarships, it does not mean that they only have N vacant places for PhD students.

3) When you apply for a scholarship, regulations vary. Say, at my institution they mostly pay attention to specific predefined points (such as scientific publication record), so Google internship won't help in this case. However, the process is funder-specific, so you'd better find out the prescribed criteria if you can.

  • 9
    joining a PhD program in CS without financial support would indicate that you are at a university you would have been better off avoiding (at least in Europe and North America). So your entire first part is inaccurate. – Tobias Kildetoft Jan 28 at 11:01
  • Why? I can name you at least one top British university (~top 10 in CS field in the UK), having a large number of self-funded PhD students from affluent families, who can afford tuition. Everything is not so black and white, places with high reputation attract everyone, including people who don't need scholarships. – rg_software Jan 28 at 11:10
  • 11
    I disagree on Phd programs are not over crowded. Competition is fierce for good positions. My uni offers 40 places out of ~300 applicants every year on average. – Koenig Lear Jan 28 at 11:11
  • @Koenig Lear, Yes, I can't speak for all institutions. Do you mean you provide 40 positions without any financial support? – rg_software Jan 28 at 11:12
  • 4
    Somebody who can get an internship at Google is going to be looking to do a PhD at a strong institution. Such institutions are always over-subscribed. – David Richerby Jan 28 at 15:45
1

In general, an internship at google has a good reputation. If it is in the desired field, it would be my choice. Especially if it is in machine learning, you are in a very good place (assuming you are not just cooking coffee ;-) )!

The most important is what you learned during the internship, and this includes cs skills as well as company culture, work ethics, etc.

You should consider which knowledge / skills you are lacking and which of the two position helps you in filling that gap. E.g. if you did nearly no research until now, the research group might be a better idea (maybe you figure out that you hate research afterwards ;-) - and/or you are gaining research skills), if you need to learn practical software development and want to know the real world, an internship in industry is favourable.

  • Um. Google internships wouldn't have a good reputation if they were just about making the coffee. (Though making coffee is a valuable skill to bring into any CS department.) – David Richerby Jan 28 at 15:47
  • @DavidRicherby How about building a coffee machine that can be controlled by voice and can learn from your ratings ("Hey Google, your coffee is too acid/too bitter/perfect")? That can lead to a PhD in HCI (Human-Coffee interaction). – user4052054 Jan 28 at 18:20
  • @user4052054 what about a camera to check the coffee level in the machine? Oh, been done already.... – Solar Mike Jan 28 at 21:40
0

What Google and Amazon do with machine learning is so far on the "applied sciences" side of things that it's probably irrelevant to your studies.

Their AI/ML projects are about throwing existing algorithms at existing data, and very little research is done in-house.

Relevant Twitter thread

  • 3
    They do research, see AlphaZero, Word2vec etc. However it's unlikely that an intern with a BS/MS in CS would land there. – Koenig Lear Jan 28 at 17:20
  • 7
    In 2017, Google had more papers accepted at NIPS than the MIT (or any other university). I'm quite sure these numbers were similar last year. – YYY Jan 28 at 20:30
  • well this is just wrong, and the twitter thread looks more like a rant than anything else. He has some good point, but he takes them way too far. Are you telling me word2vec or alphazero have not changed the field significantly? Or are they also something that he could do in 2 weeks with a random master student? (roll eyes) – Ant Jan 28 at 22:09

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

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