Computer Science PhD trends
blog.vivekhaldar.com
blog.vivekhaldar.com
Moreover, don’t believe those numbers. It is just buzz. If Google pays you 200K, it is not for your PhD, neither for your genius research. It is only about business.
I'm doing this (a PhD in English), but I wouldn't have started if I weren't able to work in my family's consulting business at the same time. Most of my peers in grad school appear to be afraid of the real world and/or diligently putting their head in the sand; now I sent prospective grad students here: http://jseliger.wordpress.com/2012/05/22/what-you-should-kno... and to similar literature on why academia in the liberal arts is an incredibly bad investment.
1. EB-1: A visa given to an alien of extraordinary ability. The requirements for this are complex but a person with a decent lawyer should be able to navigate them provided the person has access to enough people in academia and industry who vouch that this person is of extraordinary ability.
2. EB-2: Employment sponsorship. Both academic institutions and industry have a standard practice of sponsoring people who have done any form of grad. school to a path that involves a green card.
Now, sure, 1) has overly bizarre documentation requirements (like for example requiring that one print out every paper that one has written and send it to USCIS) and 2) takes way too long for people born in India and China. This of course requires reform in a fair manner that takes into account other constraints like making sure that immigration doesn't become top heavy from two nations.
The sequoia capital article talks about foreign born students getting ejected. Do we have any quantitative data on how many of them are really being ejected? ( Acquiring a H1B visa to work in the United States is not hard. When, I (a foreign born student) graduated and applied for an H1B visa, it took ten days for it to get approved. ) Have there been surveys done on people who have voluntarily left for personal reasons vs. people who left because of this issue?
1) People who have non C.S. degrees even from good schools. Once you have finished your degree, you have three months to find a job. Some people don't find a job within those three months and are forced to leave. Honestly, I am not sure what the solution to that is? Stapling a green card on their applications post graduation seems good to help an indefinite stay but what if you can't find a job in 6 months? 9 months?
2) People who don't want to stay here. Their objective is to go back home.
Actually, the difficulty varies from year to year and from company to company and from subfield to subfield. So yes, people do suffer from the status quo.
Just throwing this info here in the interest of completeness.
Actually, I do think you can apply for an EB-1 before acquiring an H-1B. I presume your concern is because of the single intent nature of the F1 visa, if so, there are people who have performed an adjustment of status whilst on OPT. Now, again in the interest of full disclosure, the only reason I applied for an H1B was because I knew that I would be applying for an adjustment of status in a few months and decided to be safer than sorry. However, I came to know later that quite a few people on OPT applied for one and didn't have problems.
> Actually, the difficulty varies from year to year and from company to company and from subfield to subfield. So yes, people do suffer from the status quo.
I am not denying there are exceptions: Some exceptions which seem justified to me include denial based on the fact that the company is a sweat shop which USCIS is cracking down on more (I am not opposed to this.). The other is delays (which can be indefinite) due to security reasons. This is especially guaranteed if you work in robotics, machine learning, biology or are a muslim etc
My point was that in general the system is not completely irrevocably broken down. Yes, it requires cleanup, a refactoring to make it simpler. Personally, I hate hate hate the H1B system and think it is one step above bonded labor, where a unscrupulous employer can fuck with you if he wishes to. Also, there needs to be considerable reform in the way green cards are allotted on an employment basis: I would love for a point based system that integrated all the different paths to permanent residency that we have right now.
Before getting into it, I'll say that I agree on policy. It is pretty dumb to send CS PhD recipients from reputable universities home, or force them to go through the idiocy of temporary work visas with dual intent to remain in the us. It makes more sense to just grant residency.
I just have a very, very different opinion about the rest of this.
First, I think that PhD programs in the sciences are generally a poor return on investment relative to other options (the professional degrees, in particular). Here's a link to a RAND study that concludes that the American allergy to STEM PhD's is rational and market-driven.
http://www.rand.org/pubs/issue_papers/IP241.html
My guess would be that PhDs in CS are a better investment than science degrees in the aggregate. However, PhD degrees in CS still have sky high attrition rates and time to completion compared to the professions, and salary and employment prospects long term for top students who could attend elite law, medical, business, or other professional schools are still better than what is offered by PhD programs.
In short, I think that any "shortage", to the extent that it exists, is market driven and rational. PhD programs need to face a reckoning, and they need to change. The RAND study makes some good suggestions, but PhD programs won't change if they can position themselves as gatekeepers to US residency. They'll just wave goodbye to the Americans who already have residency and don't have to go through this bullshit, and continue their abuse of students who have fewer legal options to reside here.
This brings me to my next disagreement - that PhD students make up the costs of their scholarship "to some extent." I think this is a vast understatement. Talented PhD students in STEM fields are profit centers for universities. The work product of a 5th year PhD student in a biochemistry lab vastly exceeds the minor stipend this student is paid. This is why PhD programs like to string out their students. It's a big reason for why they won't move to the professional model (advocated by the RAND study).
In short, I hold opposing positions here, depending on which perspective I take. Yes, PhD graduates from STEM programs should be granted residency, it's nuts to send them home or force them to jump through hoops. No, PhD programs should not, under any circumstances, hold the power to grant residency to international students. They will abuse this - they already do.
1) Even if you get a PhD and aren't shining at your field (read gulped into tenure track positions at a top 10 CS program) you're not confined to becoming a professor at a state school. This is a legitimate fear, and one of the things that has always made me frightened of pursuing a PhD.
2) Even if I end up outside of Academia, I'll likely bear little to no fiscal loss. The work will most probably be more interesting than the work I'd be doing had I not done a PhD. The job-ceiling will also be significantly higher, or non-existant.
As a semi-related question - does anyone with a background in the area know if it's easier to get accepted into a top-10 CS program having been born and raised in the States? GPA, research experience and my school's 'ranking' is a non-issue - I'm just curious if residency in the United States ever affects PhD committees in regards to admission.
That's not at all what I got out of it. I read that phds provide no material advantage and that they're really tough (in every way, including from a psychological standpoint).
>you're not confined to becoming a professor at a state school. This is a legitimate fear, and one of the things that has always made me frightened of pursuing a PhD.
Confined? Most people don't have this choice. Fear? This a dream of a large % of phd graduates. I'd add one thing that the article doesn't mention: graduate school is far less of a meritocracy than people outside realize. There are so many levels of luck involved, including finding a good advisor, department politics, finding a good project, not having your research 'scooped', funding, etc. These things are correlated with intelligence/ability, but a large number of smart people fail grad school for reasons beyond their control, and a significant % of the people who succeed were blessed with amazing luck. It's the smart and accomplished undergrads (which I take it you are) who have the most to lose.
>I'll likely bear little to no fiscal loss.
Consider the opportunity cost. Here are some reasonable numbers: a masters costs 25K, a phd pays out 150K over six years, and in the five years of working the masters makes 500K. By the time the phd is out of school the masters student is 325K ahead of the phd student, and his work experience/career is probably further (meaning his salary will probably be at least as high). Also, the masters student lived more comfortably in those five years.
> The work will most probably be more interesting than the work I'd be doing had I not done a PhD.
I have seen very few industry jobs that require a phd, and I work in a field dominated by phds. The exception is industry research, which is great but isn't open to most phds. Most phds don't be doing anything resembling their research two years out of grad school.
I don't mean to make a phd sound totally crappy. There's a lot you can get out of it, it teaches you how to think and organize projects in a way that's hard to find outside grad school. You'll meet smart, motivated people. You'll be able to interact with professors as almost-peers. But consider a research-oriented masters, it's what I did and I consider it a happy medium.
I don't think GP meant it as "stuck as an assistant professor at UT Austin" -- more like "choosing between becoming an adjunct professor at University of Nowhere-You've-Ever-Heard-Of and not having a job."
I'm not a CS PhD, but I'd be astounded if they pay tuition.
I knew PhD students in a variety of fields when I was doing mine, and everyone received money from the University rather than paying tuition.
You're point about opportunity cost stands though.
If you measure how much money someone will make in the next 10 years from the point they decide whether to pursue a PhD or not, yes, you are correct, the ones who do not pursue one will have lots more money.
However, you can live a middle class lifestyle while gettings a PhD (OK, maybe "lower middle class" here, but still very comfortably, if you don't have dependents to support) and also after getting a PhD, which is the same result as if you didn't pursue the PhD.
To me, that is more relevant than a raw amount of money.
I care what class I'm in, i.e., (a) starving; (b) poor; (c) middle class; (d) rich. I don't really care _within_ those categories.
As a sidenote, I think getting a PhD increases your likelihood of getting an otherwise unlikely outcome in the sense of career success/advancement, getting rich, etc. (unless you choose academia). I mean, over the course of your career, you could really leverage your PhD, or you could not. In theory (and probably in practice most of the time), it won't hurt to have one, career-wise.
To give you another counterpoint, as a person who dropped out of a C.S. PhD and went to industry. Sure, it won't hurt to get a PhD. Getting rich unfortunately is only weakly correlated with technical ability beyond a certain point. 1) You are not going to get rich purely as a salary (wo)man unless you are lucky enough to be an early stage employee (which I don't see how the PhD or otherwise helps you that much. ). 2) To found your own company and leverage your PhD skills seems tempting but rarely if ever do PhD's have research that can be converted into successful industry products. Boston Dynamics is one of those rare examples. However, these opportunities are not closed to off to people who while they may not have done research in the field and know every other implication of certain strategies but have great contacts and know how to get things done.
I'm just interested in your experience of doing just that. I've found very little reference material on it, except some comments by professors and people currently working in industry that people are courted by companies once they finish their qualifying exams.
This may be off topic, but what's your experience with this? Were you courted before you decided to leave? How far did you get? What affected your decision? What was the salary difference? What do you do now? Was it worth it to even go for a little?
I couldn't let my advisor down. After sticking with me for two years, I would have felt horrible if I had left him after my masters. I did not _expect_ this going in, and I planned to have "leaving after my masters" as a real option.
To be clear, I still would have left if I had had a good reason. Ultimately, I wanted to stay. But if it had been borderline, I would have still stayed. And "borderline" for grad school may very well be "I'm tired of killing myself with overwork, I don't really feel like doing this shit for a few more years. But I could just suck it up and keep going, since I'm already like half way there." Grad school can be pretty crappy.
I definitely could have turned summer internships during grad school into full-time work, though, and gotten a well-paying engineering job.
There are tons and tons of companies that come to a school every year. If you have a linkedin profile or some sort of web presence, you will be courted. It doesn't necessarily happen once you finish quals or whatever, but it will keep happening.
> This may be off topic, but what's your experience with this? Were you courted before you decided to leave? How far did you get? What affected your decision? What was the salary difference? What do you do now? Was it worth it to even go for a little?
I do data science (machine learning, scalability engineering) stuff. Lots of people have PhDs, I presume they make a good 20% or so more than I do. However, keep in mind that they have spent 3+ years finishing a PhD so I can probably make up. Honestly, if you are going into industry, I don't think you'd need a PhD. Companies these days think it is fancy to hire someone with a big name degree but a master's degree hasn't hampered me. I miss school terribly mainly because I felt stupid there (I know how silly this sounds) and the problems were terribly hard.
I think this is basically true, but I wonder about a couple of potential exceptions:
(a) I suspect that in huge corporations (IBM being a canonical example), just having the credential can help qualify you for leadership roles (e.g. leading, say, groups of 100 to 1,000 people), if you are also a good leader/strategical thinker/do-er. And at some point, these leaders are probably getting paid significantly more, which (with wise investments) could potentially (?) be bootstrapped into getting rich. I'd be interested if anyone with personal experience can comment on any of the above.
(b) I would think that having a PhD would help qualify you for the top-level executive roles. Although, frankly, I've looked at company websites, and top-level executives with PhDs are not _that_ common, even in software companies. But yeah, I mean, there _should_ be a top-level decision maker who is _at least_ keeping a close watch on any research related to what the company does (I'm talking software companies here, not companies that _use_ software)... and the vast majority of people who have the necessary skills will be people who did PhD-level research. Anyway, this person should be able to advocate for how the company can leverage new research, and this should not just be left to lower-level (hierarchically) technical folks who have no true strategic voice in the company.
To found your own company and leverage your PhD skills seems tempting but rarely if ever do PhD's have research that can be converted into successful industry products.
Do you think this is true even of CS PhDs? I feel like in my research group, there is a real chance for any of we (students) to do this, but AFAIK I am the only one who has ever really thought about it, because my colleagues tend to keep their noses in the books and focus on narrow technical concerns, whereas I'm really a big-picture thinker. So: in my case, plenty of opportunity, just not much interest among actual PhD students.
> Do you think this is true even of CS PhDs? I feel like in my research group, there is a real chance for any of we (students) to do this, but AFAIK I am the only one who has ever really thought about it, because my colleagues tend to keep their noses in the books and focus on narrow technical concerns, whereas I'm really a big-picture thinker. So: in my case, plenty of opportunity, just not much interest among actual PhD students.
My experience is mostly of C.S./Machine Learning PhDs. Let me give you an illustration. Say you spend your entire PhD figuring out one specific problem in recommendation systems, like for example, building optimization algorithms where the error rate is 6% or so. This is incredibly cool stuff but when you go out into the real world, you don't necessarily need that fancy algorithm inorder to solve problems. Really simple stuff works and the way production code works, keep it simple stupid is an important thing!
I can't see how it could fail to decrease it, once you take opportunity cost into account. The most likely way to get rich in a technical career is doing a startup. During the five or six years you were spending on your PhD, you could have founded two startups, had them both fail, learned from the experience, taken a third shot, and be seeing your latest startup taking off, all by the time you would have been putting your PhD in your pocket and wondering what to do next.
But regarding startups... doing a technical startup that leverages PhD research (or just background knowledge), or even just a "highly specialized" software consultancy, is probably a better strategy than trying to build the next stupid app that anybody who can program can build.
One key step to me going into grad school was realizing that I could always do startups, but I could only go to grad school while I was young. (Technically you can always go, but for me, there was a very strong preference to do it "now" or do it never.) So... you can have both, but probably only if you do grad school first, not startups first.
Quant finance is an option, though I suspect once you actually got into it, you'd find making serious money that way wasn't really easier than doing a startup. Besides it's mostly a zero-sum game these days, and wouldn't you like to make the world a better place as well as getting rich?
I know of people who've gone back and done a PhD in their forties or fifties. Strikes me as probably more practical than doing it in your twenties and putting the rest of your life on hold.
The probability of being able to use your PhD thesis in a startup is negligible. Technical knowledge that you pick up on the way, sure, but that's a very inefficient way of obtaining that kind of knowledge.
Quant finance is an option, though I suspect once you actually got into it, you'd find making serious money that way wasn't really easier than doing a startup.
I don't know, I mean, I wouldn't go solo... I would join up with an established company that wants to hire. And I have definitely seen these companies recruiting CS PhDs in the last 1-2 years.
Besides it's mostly a zero-sum game these days, and wouldn't you like to make the world a better place as well as getting rich?
Might be better to make some good money, and then retire kind of young and focus on whatever else you really want to do with complete financial freedom.
I suspect that traders actually do contribute though, just like every single other sector of the economy. I mean, traders provide liquidity and also "provide" econonic information, both of which help coordinate the economy. And if high frequency traders aren't actually providing any direct benefit to anybody, we should see markets arise that disallow that kind of trading.
I know of people who've gone back and done a PhD in their forties or fifties. Strikes me as probably more practical than doing it in your twenties and putting the rest of your life on hold.
I've been told by profs that older folks (who are, by implication, settled, fully mature, and have figured out what they want in life), are a much safer bet as grad students than younger folks. So, there is something to this. But if you get your PhD that late, there's not necessarily that much time left in life to do that much with it.
Technical knowledge that you pick up on the way, sure, but that's a very inefficient way of obtaining that kind of knowledge.
I think it's counterintuitive, but I disagree. If you really want to understand the cutting edge and see new opportunities, you have to be carefully reading the research papers that are being published (and understanding them), doing a lot of critical thinking, and talking to people in the field. And it's going to take a few years. A grad student is well-positioned to do this. Anybody else who wants to do it almost might as well just be a grad student (unless they're already a professional researcer or professor, which typically implies having been a grad student).
That's an attractive idea in theory, but it almost never works out that way in practice. Your brain rewires itself over the years to match what you're doing. Unless you are very unusual, your expenses will drift up to match your income. You are almost certainly much better off to make your plans as though today was the first day of the rest of your life.
> I suspect that traders actually do contribute though, just like every single other sector of the economy.
Absolutely, they do. But it has to be past the point of diminishing returns by now. Cutting the time to move capital from a week to a day was surely a contribution to the economy. Cutting it from a hundred milliseconds to fifty milliseconds? I have a hard time believing that does more for the economy than writing a better poker bot.
> If you really want to understand the cutting edge and see new opportunities, you have to be carefully reading the research papers that are being published
Business opportunities usually arise some way behind the cutting edge. You are right of course that you don't want to fall into writing yet another cat photograph website because you don't know how to do anything else. But neither do you want to waste the best years of your life obsessing about the mathematical properties of some esoteric algorithm that ends up being no better than off-the-shelf algorithms on practical workloads. The sweet spot tends to be somewhere in the middle.
He might be not bad at managing money but almost everyone who starts earning more starts spending more. This is not bad money management, this is human nature.
Another effect is cost-of-living changes when moving from college town to real life. If a graduate earns 100k, it is probably not in a low-cost area. Taxes on this income are also higher than on a PhD income.
As a CS PhD student, I have asked this question before of professors on the admissions committee.
Once the applications come in, they sort them into 3 separate piles: US, India, and China. Each pile is evaluated separately, because they aren't comparable. e.g. grades and prof recommendations from the US matter a lot, whereas from China, it's hard to compare grades and prof recommendations are almost always fraudulent. (I don't know what differs about India.)
Then they accept the top ones from each pile, and try to overall choose the top students across all piles. So it's hard to compare "maybe accepts" from different piles, but they do their best.
I guess other countries (besides the big 3) get treated differently because they are more rare.
So, basically, the answer is no, being US based does not really help or hurt you.
I have heard that you have to be extremely good if you're Chinese, because if you're not, it's just too hard to compare you to people from the other piles reliably.
Obviously, this story is highly anecdotal. I'm sure other departments have other ways of handling the problem of compraring people from different educational systems.
It feels like the main difference here is that the admissions committee looks for evidence of high/low English proficiency. There are also plenty of things other than grades to look at -- test scores, research experience, interests that match faculty here, etc.
That said, how do you judge the English proficiency of someone from India or China, without meeting them and working with them?
They make them take standardized tests that show very basic English skills, but I think you can't really judge that well in terms of doing high-quality writing and good interpersonal communication.
Speaking english is a certainly a plus, but it seems outweighed by all the other benefits to the advisors that a foreign student brings.
(Full disclosure: I'm an American who's going to apply next year to PhD programs.)
But, more likely to be pressured by their family into a "prestigious" career that they aren't really interested in, more likely to have the additional pressure of supporting a wife and be thinking about kids, more likely to be unable to learn how to write research papers well in English (this is surprisingly hard)...
Also, I've definitely seen people from abroad use American grad school as a temporary stepping stone to get a job at a big company (like Microsoft) that will deal with the visa stuff. Not saying they were planning it all along, but it happens.
Of all the above, I think the "being able to write polished, professional, technical research papers" is by far the biggest risk factor that neutralizes the benefits of foreign students that you mentioned.
Most extremely smart academia-bound computer scientists consider that a dream, not a fear. Realize that many state schools are in the top bunch (e.g. 20 or 30) in the US. Unless by "state school" you mean "shitty state school." Like, "X State" where X is not actually a U.S. state (no offense, but those annoy me).
For a professor, the main difference in quality of the school you work at is the quality of your students. This matters a lot, research-wise, because most profs that do a lot of top-quality research farm it out to their grad students. e.g. with 8 good students, you can increase your productivity about 8x.
Post-docs are safer jobs, because you know what will happen at the end of them (you get kicked out and find another job). Fewer people going straight into tenure-track means fewer people spending years at an institution, not getting a job (because they lost the race) and starting elsewhere. Sounds very healthy to me!
So 23.9% are in other parts of academic computer science, mostly Postdocs. another 3.6% go to positions (unspecified) in other academic departments.
His 47% is "North American, Non-Academic, Industry". It doesn't break down research vs. development jobs, which is probably smart.
Another interesting thing is that it's broken down (Table D4) by specialty - so for instance, two PhDs in HCI left the US from the 2010-2011 class, and the specialties with the highest percentage leaving the country after graduating were "Information Systems" and "Networks".