For anybody interested in the computer science PhD process, a good complement to the featured article is Philip Guo's "The PhD Grind" [1]. It's a long (115p, but with big text) blow-by-blow account of the author's PhD at Stanford in the late 00s. It's neither sensationalized nor sanitized, and it's weirdly engrossing for an account of PhD research. I read it before starting grad school and I think it helped my "meta" PhD skills a lot.
Applying to PhD Programs in Computer Science (2014) [pdf]
41–50 of 106 posts
Re: Applying to PhD Programs in Computer Science (2014) [pdf]
#42> "...You need to know why you want a Ph.D. You need to have vision and ideas and you need to be able to express yourself..." I've got to stress that this advice is really important. Especially for people who are considering going into grad school because they're on autopilot or think that it's just a natural continuation of your schooling. That's an easy way to think, but it's not the right way. You can easily lose…
> Go into grad school and pursue a PHD because you've got a burning desire to turn one of your good ideas into a breakthrough.
I don't think many supervisors will let you do this. 1) you'll need to find someone who has funding and already works in the area of your good idea 2) people overestimate how much they can do during a PhD 3) you probably aren't aware of the field coming out of undergrad 4) not a lot of theses will ever be read outside of the examiners, let alone be a breakthrough.
I'd say do a PhD if you feel like it, you can work for years on mundane tasks that likely wont culminate in a breakthrough, aren't prone to depression/stress, and don't care about money.
Re: Applying to PhD Programs in Computer Science (2014) [pdf]
#43> "...You need to know why you want a Ph.D. You need to have vision and ideas and you need to be able to express yourself..." I've got to stress that this advice is really important. Especially for people who are considering going into grad school because they're on autopilot or think that it's just a natural continuation of your schooling. That's an easy way to think, but it's not the right way. You can easily lose…
What would you do to get a good idea? I have an area I'm really interested in, but it just seems like any ideas I have could have the potential to be a dud. Coming up with an idea that you'd be willing to commit several years to just seems incredibly hard. I'd imagine that working at a real company isn't really the best occupation and place to generate new ideas unless what you're interested in is more practical. It…
> I'm from Europe so I wouldn't be in debt or anything
Do a Masters and it will help with all of your questions
Re: Applying to PhD Programs in Computer Science (2014) [pdf]
#44My feelings are a little different. If you feel like doing a PhD and enjoy doing research, you should do it. A PhD above all should be fun, and is thus not a waste. It's only a waste if it's not fun. There will be extended periods of intense struggle, and those parts are not that fun. One has to accept those. But, doing cutting edge research with experts is an incredible opportunity and I'd encourage those interested…
Agreed. As you noted, you don’t have much to lose by giving it a try, and you can always drop out and get the MS (like I did).
Not much to lose... except four to six years of your life. If you think of your working life, which has maybe 45 years, four to six years is a fairly major cost.
That's a fairly high cost even if it's fun. It still may be worth doing, and the fun may tip the scales, but... "don't have much to lose" is I think inaccurate.
Re: Applying to PhD Programs in Computer Science (2014) [pdf]
#45Earlier quoted context omitted.
What would you do to get a good idea? I have an area I'm really interested in, but it just seems like any ideas I have could have the potential to be a dud. Coming up with an idea that you'd be willing to commit several years to just seems incredibly hard. I'd imagine that working at a real company isn't really the best occupation and place to generate new ideas unless what you're interested in is more practical. It…
> I'm from Europe so I wouldn't be in debt or anything Do a Masters and it will help with all of your questions
I'm in the process of getting my master's degree, but so far it just feels incremental to the bachelor's. Does the thesis make a huge difference?
Re: Applying to PhD Programs in Computer Science (2014) [pdf]
#46Earlier quoted context omitted.
easiest way to jump tiers is networking i.e. person to person conversations with PIs. if you're able to represent yourself well and describe your own research articulately and authoritatively it'll make up for a great deal of the gap between you and students from "elite" schools. unfortunately this networking happens at conferences which less resource-rich schools can't afford to send undergrads to. my hack: when you…
This comment is extremely helpful to me right now and I'm sure may be to many others as well. Thanks for writing it. If I may, how did you prep for those chats? What did you .. say when you got there? Read some of their papers and bring followups to their conclusions? For anyone else reading, I want to echo the sentiment here 100x. I'm currently working closely on a research project with a CS prof who is one of the m…
>If I may, how did you prep for those chats? What did you .. say when you got there? Read some of their papers and bring followups to their conclusions?
like I said the way to approach this is to try to identify with the person you're pitching to and address their needs/goals.
The most obvious thing is to familiarize yourself with their research because they have research goals. But honestly you don't need to go overboard on this - no one that's not their student yet would be expected to know the research in depth. Pick 3 of their most recent papers, read the abstracts and conclusions and come up with vaguely interesting questions. Again they don't need to be deep they just need to be better than "what learning rate did you use" and more sophisticated than "why did you pick Adam over rmsprop". Obviously extra points if have your own perspective on something (even if it's not earth shatteringly novel). For example my new group works on distributed ML and on my first call I said that I'd always wondered about distributed inference (ie one test sample forward pass distributed across multiple nodes). I didn't have an answer or solution to the problem but it was something they hadn't considered.
The better thing though to discuss is new potential funding directions that you would enable them to avail themselves of. Figuring this out is admittedly tough if you're an undergrad because you don't know what you're good enough at that you could get funding. Again for example in my case I have some computer vision experience and so now my group can plausibly apply for grants at the intersection of distributed systems and CV. I didn't quite discover this intentionally but early conversations broached it organically and I recognized the opportunity to include it in my "pitch".
But even if you don't have some kind of skills alignment like that, the most important thing to stress is that you can independently start and complete projects and you can write/present academic work. That means you can speak in jargon, you know how to display results (plots, comparisons benchmarks, etc), and most importantly you can concisely explain what you've done. I had an unpublished lit review that I sent over as a writing sample.
This is getting a little long so I'll just say something everyone should always keep in mind: you're trying to build a rapport with a person. No matter how smart, successful, elite they are in their area they're not robots they're fleshly people first. Being friendly and funny and getting to know them outside of their research helps immensely (during the meeting not before - that's creepy!)
edit: sorry also another thing you can do is prep by speaking with a grad student and asking them good questions about the research, group, goals, PI, etc. this'll get you comfortable speaking in terms that the PI is certain to relate to.
Re: Applying to PhD Programs in Computer Science (2014) [pdf]
#47Earlier quoted context omitted.
This comment is extremely helpful to me right now and I'm sure may be to many others as well. Thanks for writing it. If I may, how did you prep for those chats? What did you .. say when you got there? Read some of their papers and bring followups to their conclusions? For anyone else reading, I want to echo the sentiment here 100x. I'm currently working closely on a research project with a CS prof who is one of the m…
>If I may, how did you prep for those chats? What did you .. say when you got there? Read some of their papers and bring followups to their conclusions? like I said the way to approach this is to try to identify with the person you're pitching to and address their needs/goals. The most obvious thing is to familiarize yourself with their research because they have research goals. But honestly you don't need to go over…
This is very helpful! Thanks for taking the time to reply. "The better thing though to discuss is new potential funding directions that you would enable them to avail themselves of." This is something I haven't considered before and will be mindful of now.
"No matter how smart, successful, elite they are in their area they're not robots they're fleshly people first." Hah, +1. I actually wasn't aware of my current group's "status" until well after my initial contact. I just thought our research interests matched perfectly and I was excited to learn about what they were doing. Enthusiasm goes a long way.
Re: Applying to PhD Programs in Computer Science (2014) [pdf]
#48> Some people never do research again after completing a Ph.D. For such people, the Ph.D. was largely a waste of time. Is it? Some of my friends went on to start successful companies after getting their PhDs. I'm a bit envious of them actually, I didn't have that option as my research was on an obscure topic with zero commercial potential. > The Ph.D. is a tremendous opportunity. You get to pick an advisor in any res…
I think the value of the whole booklet is "here's what we think about our PhD program," not necessarily "what place do PhDs really have in society?"
They could be wrong about the latter, but if they believe it strongly at CMU, it's good (especially for students who might go there) that they say so.
Re: Applying to PhD Programs in Computer Science (2014) [pdf]
#49I don't work in CS. I was surprised when I watched this talk[1] by Michael Stonebraker to learn of the "Diarrhea of Papers" > A PhD applicant who wants to get a good academic job should have ~10 papers > A tenure case should have ~40 papers That alone would push me away from a PhD in CS. In economics, it's closer to what he said was the case when he was new: > I got tenure from Berkeley in 1976 — With about 5 papers…
There's a bit of a counterpoint to this: with papers, it's easier to point to your body of work (peer-reviewed and, if you're good/lucky, peer-understood-and-approved) as clear evidence that you can do research. Without papers, or with very few papers, that evidence has to come from somewhere else. My understanding is that somewhere else is usually recommendations, from your advisor and other people in the field. Som…
I appreciate your perspective. The people on the hiring and tenure committees as far as I can tell usually don't/can't evaluate quality directly so they instead focus on proxies like the total number of papers and what journals someone published in. I'm fairly against total publications as a proxy but I hadn't considered how it fared against these alternatives.
Personally, if I saw someone published a ton of papers during their PhD, I'd question the quality and importance of their research. Quality tends to decrease with quantity in my experience. And it appears too easy to fake importance. In my field I watch out for vague words like "novel" and "fundamental" in paper titles as evidence that the work is probably overhyped. I'm fairly well read in my field and I can sense the disappointment in people when I tell them that their "great" idea was done before, particularly when it was done better and/or a long time ago. (I take no pleasure in doing this. I just wish people spent less time reinventing the wheel.)
I think I could have had 10 papers during my PhD if I focused on easily-published things and "salami-sliced" my papers. But I focused on what I thought was more important and made the scope of each paper larger than normal because it saves readers time.
Proxies that are characteristics of the researcher might be better (or at least should be considered), like the willingness of the researcher to put "skin in the game" so-to-speak. I'd have a lot more confidence in research conducted part-time by someone on their own money than someone using another person's money (which has much less risk). Taleb has written about this as I recall.
Re: Applying to PhD Programs in Computer Science (2014) [pdf]
#50Most Comp Sci PhDs work from home anyways once they’re in.
If by work from home you mean live from the office then yeah, sure :)