Hacker News new | past | comments | ask | show | jobs | submit login
What My PhD Was Like (jxyzabc.blogspot.com)
217 points by ingve on Feb 28, 2016 | hide | past | favorite | 119 comments



Look, a PhD is for one thing, and one thing only: Teaching you to be a researcher. You're basically an apprentice researcher, learning from your advisor and others in the field the current state of affairs, the language they use, the questions that remain unanswered, and the methods you can use to answer new questions.

A PhD is not proof that you're smart, or that you cannot get a job, or of anything else (positive or negative). It just proves that you've demonstrated an ability to do original research, and to survive a system that is often unfair in numerous ways.

It also implicitly assumes that you're young -- and better yet, single and without kids. In that sense, it's sort of like a startup, in that it consumes as much of your time and energy as you're willing to give it.

I went into my PhD program knowing that I wasn't going to be a researcher. I went in planning to learn new things, to meet new people, and to maybe try some new directions in my career. In the end, it took a very long time (11 years), cost my family and me a lot of time and money, and didn't directly affect my career. And yes, there were lots of tears along the way, as well as many angry words.

I'm very happy that I finished it, rather than give up in the middle (which I considered on many occasions). I have a huge sense of accomplishment. I learned a ton, that's for sure. And we had an adventure living in Chicago, where I met many new friends.

But I tell people, very seriously, that if they're thinking of doing a PhD, they should think very hard before starting it. It's a huge commitment of time and money, and especially in the computer industry, there are ways to do research nowadays that don't require the effective vows of poverty and servitude associated with graduate school.


> But I tell people, very seriously, that if they're thinking of doing a PhD

I simply tell them not to do it--in Computer Science--if they are genuinely good at getting things done (for some mediocre people good at playing a certain type of a game, grad school can be an okay career path), unless (1) they need a student visa, or (2) they definitely want to go into academia (which generally only makes sense in a top 5-10 ranked school).

To me the litmus test is now this: do you need a particle accelerator (or equivalents) to conduct research in your field? If not, don't do a PhD.


I like the particle accelerator litmus test :p Also .. totally agree with the top 5-10 school. Universities do not typically "hire down". I heard an "unofficial list" during a faculty meeting my last year .. I was horrified!

Here is the big problem IMHO: by the time you graduate with your PhD in a 6-8 year program, you are either close to or in your 30s (best case). You just spent a decade investing in a research career. Most scientists will tell you that research is not being supported the same as in the old days. I can't say about govt grants but industrial labs have definitely changed in the last 10 years (I'm thinking of IBM Research and Microsoft Research). There have been several lab closures in the recent past: MSR's Silicon Valley, Nokia Research in the Bay area, Intel Lab-lets (Berkeley, Pittsburgh, etc.). So basically, supply is steady and demand is decreasing. This generally doesn't bode well for the prospective PhD candidate trying to make a career in industrial research. The bar simply keeps getting higher to get one of these "coveted" positions.

While it is true that PhDs can (and do) get hired at places like Google, Facebook or Wall Street, think about the almost decade spent on training that not only are you not using, at least a chunk of the technical content will be obsolete by the time you finish! Insane!!

To succeed with a PhD, you need to be a great seller. This is sort of the same skill you need as an entrepreneur or a specific type of middle-manager (dreamer? not sure this has a clear job title). I always thought it was about the technical challenge of the work or the quality. I sadly have to come to realize that the marketing (the intro to your paper) matters as much as the rest.


You can gain a lot of education working on a phd at a good school, even past your masters, and that can help if you are doing more sophisticated stuff. I found my phd helped a lot in my dev work at microsoft and google. Maybe someone with a lot strong BS or MS could have done it, but I benefited from my 8 years at grad school! I'm kind of embarrassed to admit how long it took, but I did finally finish :-)


That's very short sighted. There are tons of industry positions in Computer Science that pretty much require a PhD if you want to be a competitive applicant. Research isn't about expensive equipment, it's about time. Sure, there's a lot of CS research you can do in your basement for small amounts of money (depending on the field), but you're never going to compete with people who spent all day every day doing research.


This is so true. Working on a relational database, especially a big team system, most people I have worked with did have phds.


>To me the litmus test is now this: do you need a particle accelerator (or equivalents) to conduct research in your field? If not, don't do a PhD.

As someone that did research outside of academia my litmus test was:

Do I need a mentor to train me to be a great researcher? After doing research on my own for a while I discovered that yes I needed such a mentor. I optimized my search around finding the best mentor possible.


What did you find were the reasons you needed a mentor?


1. Better communication in terms of writing and presentation (for my first talk I did three practice talks a day for a month with constant feedback and notes).

2. Understanding which problems to work on (Prior to grad school I spent over a year on a problem, when I published a solution I discovered that there was very little interest. I want to build things people use.).

3. Developing better research methods (How to approach and understand a problem. How to think about solutions and measure them.).

4. Honest, critical, and in-depth feedback on all of the above.


Computer science is a broad field and not all of us are doing our PhDs in software engineering. There are some things which you just can't do in the industry yet, since the technology is not at a point where it is commercially viable. The question you really need to ask is "will this project make money in 5 months?" If the answer is no, and you currently have no credentials, then it'll be a very though sell...


A friend of mine in my program kept asking me why I was doing a PhD. He called it an entrance card to the academic world, allowing you to apply for certain types of jobs. He kept wondering why I was subjecting myself to so much work, and difficult relations with my advisor, when I had a great career going beforehand.

So yes, if you're not interested in doing academic research, then a PhD is a pretty hard sell.


> To me the litmus test is now this: do you need a particle accelerator (or equivalents) to conduct research in your field? If not, don't do a PhD.

This advice would apply to all mathematics research.


A PhD gives you a lot of freedom (time) to work on something you consider interesting. It's not without other downsides (stressful), but it's a trade-off worth considering. Also, in Belgium where I'm from (but I believe more broadly, Europe), a PhD gives you slightly inferior wages to a junior position, so you don't lose out as much as in the US.


There's a world outside of CS, you know?


Largely agree, although, working for a PhD can have useful positive outcomes.

My own story follows:

I started the PhD in a UK university in the mid ‘1980s. I started it because it was the simplest course of action (staying at a comfortable uni) not because I wanted a career in academia, and I was fascinated by the subject (building a frame-based knowledge representation to support a first-order logic reasoner). Two other grads started at the same time, working on different aspects of the same project to get their PhDs. In retrospect it was a big mistake. The project devolved into uncontrolled software development (which I enjoyed) but we were in fact spending all our time building research tools (in C!) rather than using them to do research. Our mutual supervisor had no particular team leading experience and given that it had become a group project, we lacked the discipline of a proper software development lead. Our supervisor abruptly resigned in the third year to move to a new Uni and he didn’t offer to take us with him. I decided to cut my losses and took a research-programmer job in a different department of the same university. I used my spare time to write up my thesis to the extent that I could and submitted it as an M.Phil (so I did at least get a qualification from the whole experience). I then took a job at the advanced techniques department of a large UK engineering company, where I was introduced to Lisp-based tools that would have let me complete my PhD coding in months rather than years. And have worked in industry ever since.

My message to prospective PhD students is to do the PhD for the right reasons. Make sure that you understand the approach your supervisor will want you to follow and check with their other students before taking them on. Ask what their track record is, in terms of whether their students complete the PhDs. Make sure you understand what tools you will be using, and whether they are fit for purpose (i.e., for getting research outputs), and whether there is a supportive community of users in the facility.

The plus side is that I had a great time socialy, and learned software engineering the hard way, which led to a good job offer. But 25 years on, in industry, my actual qualifications are now irrelevant to my career development and progression.


> Look, a PhD is for one thing, and one thing only: Teaching you to be a researcher.

I disagree - there's tons of skills you learn in the process which are useful in other contexts - the most important, in my opinion, is learning how to write well and communicate complicated ideas to others.

If you go into industry, then depending on what you studied and what kind of job you have, the knowledge of a given field (networking, graphics, compilers, computer vision etc) can be extremely helpful.


Two good reasons but (especially outside of the hard sciences) a third reason is possibly even more relevant than supporting a research career or a teaching skills. That "Doctor" title is prestigious. Your employer can pay you more, you can get higher positions as e.g. a consultant on projects, you will automatically get more of a certain type of respect and get listed to more. I am thinking of a management consultant type career here, but I suspect it is reasonably universal. Hell in some cultures I imagine it can even help you get a better choice of marriage partner. (I know it is a little bit fake, but often people's perception is more important than reality)


Academics write well? Have you heard of the word 'academese'? Honestly, I suspect a PhD brainwashes you into becoming a bad writer. These days I'm impressed by any academic who's not a horrible writer.


The "academese" tends to be worse in the softer fields, and isn't as bad in computer science. A large part of my job is helping students express themselves clearly -- it's part of learning to think clearly. That's not to say that all of our papers are shining examples of clarity, but they should aspire to be so.


Most academic writing isn't meant for public consumption--papers are communications between researchers, so the jargon etc. is a more of a convenience than "bad writing". Good researchers can talk very plainly about complex things, though this usually happens in talks and conversation.


Totally agree. The reason is simple: you can't count on the goodwill of your audience, you have to write in a defensive manner expecting an adversarial audience (at least the reviewers, probably the people you have in mind when you write your paper; but possibly beyond). See this: http://chronicle.com/article/Why-Academics-Writing-Stinks/14...


There are several types of academic writers, for sure. Some are amazingly clear in their writing. Those people are fantastic, and really do help to advance the spread of new ideas.

But some (many) use long words and convoluted sentences, seemingly for no reason other than to make their reader feel less intelligent. I know someone in the latter category; while he's a very bright guy, I haven't the foggiest idea what he's saying in any of his papers.

Which does indeed demonstrate that a PhD doesn't necessarily make you a great writer. I had been writing professionally for a decade before I started my PhD, and this had far more to do with my clarity of writing than any feedback I got in grad school.

And of course, don't forget the gibberish, automatically generated papers that were accepted: http://www.ibtimes.com/fake-research-papers-how-did-more-120...


Well, give me academese over patentese anytime :)


Being able to write something coherent quickly about a very technical subject is really, really hard. If someone needs to use "academese" to get there, power to them.


The quietest part of the Physics library at the university I went to was where the PhD tomes went to never be read ever again. Some unfortunate professor type of person had to read these turgid things once but that was about it, no wider audience. Why write for one to two people? That is what happens to most 'thesis' efforts, nobody reads them. Ever. I doubt anyone on here has plans to read a thesis or two by teatime, or to stock up on a few for some holiday reading. Real, valuable information is on the internet, peer reviewed by people in things like 'comment' threads and endorsed by 'links'. None of those tomes of dead tree in the Physics library made it to the internet, they are lost rather than 'search engine optimised'.

I did some work helping Msc and PhD students many aeons ago and I felt that people hid in academia at that time because there was a recession and few initials after the name were better than a spell of unemployment for those with daddies that could afford it.


Typically the thesis is based on a series of peer-reviewed papers submitted by the PhD student during grad school. These papers are actually read by people (at least for PhD students in good programs who will need to submit to good journals). The thesis is just proof to your thesis committee that you can write a comprehensive treatment of your area, and to put all of your results in one place for their convenience. It doesn't matter if only a few people read it. It is a little strange to store them all in a library somewhere, given that they won't really be read again, but I suspect this is just a carry-over from a different time when this made sense (eg. before Google made it so easy to find all of someone's papers).

The above is based on experiences in STEM, probably things are different for liberal arts people.


>the most important, in my opinion, is learning how to write well and communicate complicated ideas to others.

you also learn this while working for a company, which has the added benefit of not treating you like an indentured servant.


>But I tell people, very seriously, that if they're thinking of doing a PhD, they should think very hard before starting it. It's a huge commitment of time and money, and especially in the computer industry, there are ways to do research nowadays that don't require the effective vows of poverty and servitude associated with graduate school.

There are!?


> there are ways to do research nowadays that don't require the effective vows of poverty and servitude associated with graduate school.

What would be some of these ways in your opinion?


Any of the big companies -- Google, Facebook, and Amazon come to mind -- do tons of things that even 10 years ago would have been considered cutting-edge computer science research. Whether it's OS changes, algorithm development, database technologies, UI testing, or data science, those companies (among others) are working with far larger data sets than most university researchers can ever touch. Plus, they don't have to deal with the sorts of ethics committees that universities have in place.

On this last point, I should mention that my dissertation research involved the creation of a collaborative Web application for people developing agent-based models. The IRB (ethics board) insisted that anyone who would want to use my Web site would need to fax a permission form to the university office. You can imagine how this would have killed any use of my site, and thus of my dissertation work. It took several months of pleading with them to get them agree that usage of the site wouldn't require a faxed form. Doing that same sort of work at Facebook et. al. would have been a non-issue.

And yes, there are ethical and privacy issues associated with Facebook, Google, etc. I'm not ignoring them, merely pointing to the abundance of data and opportunities they have at their disposal.

There are still advantages to a university laboratory, including the cushion that you have from real-world profitability. There's something nice about being able to spend months or years looking at a problem, without having shareholders looking over your shoulder. But universities have their own politics and problems, and they're far from the pristine labs we'd like to imagine.


"Any of the big companies -- Google, Facebook, and Amazon come to mind -- do tons of things that even 10 years ago would have been considered cutting-edge computer science research."

Absolutely true, but is it possible to work in those places without a PhD?

One path I can imagine would be to devote 2-3 years to independent study, attend appropriate conferences and get to know the researchers on a professional and personal level. Build some respect on your own dime before trying to move to professional work.

No idea if that would work, though - whether there are avenues for an amateur-but-devoted scientist to contribute to research, or whether that research would break down any barriers.


I know someone who was doing a PhD on neural networks, and interned at Google during it (a research group involving NNs). He was told by that research team that they wouldn't hire him without his completed phd. He cut his phd short to a masters and got a job at Google anyway, but unfortunately I can't remember whether it was on that research team, or a generic software-engineer position (either way, the internship was the foot in the door).

You can certainly do good independent work, and write and publish papers if you want. If you can prove your conpetence, why shouldn't you be hired? At least, not every employer is going to have a major bias against those without PhDs.


Indeed. Look at the institution best paper awards: http://jeffhuang.com/best_paper_awards.html#institutions

There are several companies (and not just universities) in the top 20. Microsoft leading.


But most of the people in the 'Research' departments of companies will have PhDs. For example, see the list of full-time positions at MSR [0]: both 'Researcher' positions require a PhD (Researcher - "Successful candidates will have a Ph.D", Post-Doc Researcher - "Applicants must have completed the requirements for a PhD, including submission of their thesis, prior to joining Microsoft Research"), and it sounds like it is helpful for some of the 'Technical' postions (Research Software Development Engineers - "RSDEs are strong software developers with a Bachelor’s Degree, and often a Masters or Ph.D. in computer science"). There are 'Non-PhD Researcher Opportunities', but only in India and China.

[0]: http://research.microsoft.com/en-us/jobs/fulltime/default.as...


I'm at MSR in China and I have to say that, even though there are a few researchers without PhDs here, they are not that many.

And having a PhD and a body of work does help a lot in the interview cycles.


Oh, how could I have forgotten Microsoft?

Microsoft Research has a lot of amazing people working there, doing very impressive stuff.


Hello, I'm a PhD doing agent-based models. Is your site (of your dissertation) still working? Can you share the resources? (I checked your personal site but couldn't find it).

Thanks,


It's at http://modelingcommons.org/, and is for NetLogo models. (Yay, NetLogo!)

It's far from perfect, and I really need to do some maintenance on it one of these days. But as things stand, I have the largest public collection of agent-based models. One of my committee members told me that another research group had done something similar, but with a multi-million dollar NSF grant. That certainly made my day!

Let me know if you have issues/bugs/whatever!


Thanks for the resource, however, i work very little with NetLogo. Very nice reference though. Keep up the good work.


I hope you don't mind asking me what field your PhD was in?


Oh, I'm open to any and all questions!

I did it in a program called "Learning Sciences," which is a mix of computer science, cognitive science, and design, all for educational purposes. I studied at Northwestern University, just outside of Chicago.


Could you say a few words about your post-PhD career? I see from your profile that you are an independent consultant. How easy/difficult was it to get gigs? Did the PhD help/hurt in your opinion?

Also, curious if you have worked as a regular software engineering employee? It is kind of difficult to be in your mid to late 30s as a fresh software engineer when many people your age have more years writing production code and being a team lead/manager. Curious if/how you dealt with that?


Well, I've been a consultant for 20 years. That started before the PhD, and has continued since then. And while my advisor didn't know about it at the time, I was also doing some consulting work during the PhD, so that my family (wife + 3 kids) could afford rent, food, and health insurance.

My ability to get gigs was always pretty good -- but it really skyrocketed about 2 years ago. That's when I decided to explicitly pursue training in a few technologies (Python, Git, PostgreSQL, and Ruby) rather than "doing whatever they want, using whatever technology they want." That specialization, and my focus on training, has done wonders to my career; I'm earning more, having more fun, and doing more interesting work (I think) than when I was looking all over for development jobs.

I still get to do software development, but now I do it for things I want, or when researching new topics for myself. I have a few SaaS ideas that I've been thinking about, but haven't had time, between writing books and doing courses.

It's hard to say whether the PhD has helped me much. On the one hand, I put it on my ebook covers, since I figure it'll impress some people enough to buy the book. And to be honest, I have used (and continue to use) many of the techniques I learned in the LS program in my lectures, which I do think help to make me a more effective trainer. However, I think that people hire me not because of the PhD, but because I have built up a reputation -- and once you do a few successful courses for a company, they're very likely to hire you back.

I did work as a regular employee for a bunch of years: I worked at HP's medical products group (spun off as Agilent, and then sold to Philips, both after I was there), and then for Time Warner. When I moved to Israel in 1995, I figured that this was a good chance to try my hand at consulting -- and I haven't had any regrets since.

It's a sad, but true, state of affairs in the computer industry, in that people in their mid-30s who are qualified and bright have a hard time getting jobs. I'm fortunate in that I haven't had to deal with that. (And I'm now 45, so I'm definitely over the hill in terms of getting development jobs.) Consulting has provided me with something of a cushion there.


I know age discrimination must happen (even in software), because people report it, but I'm about 50 and maybe I'm fortunate but I and many old ex-microsoft friends haven't had trouble getting jobs here. I get pinged multiple times a week to interview. If you are a dev in Seattle, can really code, and can pass those famous 1 hour design & code interviews in c++ or java, there are endless jobs in Seattle (hint: always try a hash table :-)).

I get that you can't just move to seattle or some big city with tech or wherever for a job, you might have a family, a life (it's fairly expensive here), but if you can't find a job as a dev, you are probably living in the wrong place.

You have to practice before interviewing. If you don't know c++ or java or js, practice on your own until you proficient. Just get out there and try. I feel so sad when i read one of the endless stories about someone who got laid off by a stupid evil company (like the recent disney people) and training your replacement, that would be unbearable. I have been fortunate to find companies that honestly did have thousands of job openings, and did hire tons of us citizens but also we hired thousands of people outside of the us of a too.


I think it is great that Jean (and Philip) both wrote these up. But, to be honest it 100% depresses me to read them. Graduate school has not been easy for me and I had no first author paper submissions for 5 years and ended up having to being part time for 3 years to support myself and my family. (My research is well in hand now. I finally submitted my first paper and have 2 more going out this calendar year. Sometimes it takes a while to find your passion/research area.) Given my family situation I am not able to do internships and in many ways I don't want to. I am going to go back to coding now because this is too depressing to read.


I feel that the US PhD system is extremely bad not for students, but for human beings. Other countries like Australia or Germany have a much more stringent timeplan - maybe 4 years, usually 3 years. I often talk to researchers from the US and it's always 7, 8 years spent, with maybe one publication, or even 0 - the time is spent interning, or drifting between departments, taking various university courses that really lead nowhere (like OP), or drifting along without a real project, or you're the cheap work monkey of the supervisor (the last one seems to be extremely common in the US).

I think that's the universities' fault - Australian unis force you to have a project proposal after the 6 or 12 months mark, and then you have about a year or two to do that project, then you have to write. There is very little time to do courses, and if you take a course, you lose one out of 3 possible time extensions. If you don't make it in time your student visa will expire (I think maximum is 4 years?) and you have to leave the country. If you're a citizen your scholarship will end at the same time. And these are good scholarships, usually about $500-700 per week.

My PhD experience was nothing like OP's or yours, much more focused, much less time wasted, much more publications too! I think our bioinformatics group's record was 12 papers in one PhD (3 first author). I had 2 first author and a few more middle author papers, one of them in Science.


The primary output of a PhD is not a good project -- it's actually the European system that's geared towards quick cheap labor for the PI's idea. Instead, the output of the US system is researchers capable of forging their own path. No surprise Jean landed a professorship at CMU after all this. That's useful for anyone in a thought leadership role, e.g., PI/CEO/CTO. (Though yes, for related reasons, US CS systems PhDs probably also have a higher impact factor, e.g., they're actually implemented as is discussed in her post.)

As for the bean counting: her first-author PLDI best paper is the field's equivalent of a Nature paper, and she got one in her first year.


    geared towards quick cheap labor for the PI's idea. 
I am not sure the phrase "quick cheap labour" really gets to the heart of the matter.

The real reason why most PhD students work on their supervisor's ideas is different. Most students would drown in the ocean of possible ideas, would simply fail without strong guidance from an experienced supervisor. It's probably best to see a PhD as a warm-up exercise for future research. The ocean of previous research one must master before being able to do original research in mature fields is simply so large that most students can't handle it on their own. Supervisors would love PhD-students who are intellectually mature enough to come up with their own ideas and execute them to a publishable standard. In my experience, at best 1 in 5 PhD students can do this.

I used to think my own PhD work was 100% self-invented. But now, having supervised a lot of my own students, I see clearly how my intellectual mentor gently nudged me in the right direction at all critical junctions where I would have gone astray otherwise. Without him, I too would have failed.

US PhD typically last a lot longer than UK PhDs, so comparing the two is difficult. I'd say UK PhD+postdoc is a probably comparable with an US PhD.


> it's actually the European system that's geared towards quick cheap labor for the PI's idea

Hmm I thought it was the other way around. In the UK PhD students rarely do work for their supervisors like teaching undergraduate courses, which seems to happen a lot in the US. When I started my PhD in the UK I was told I needed to come up with my own project from the start. I only did one paper that was sort of related to my supervisor's project, and I only did a hundred hours teaching over my entire PhD, and that was voluntary.

The other big difference between the US and UK PhD approaches is that in the UK a PhD student doesn't do any courses. Day one on the PhD is day one working towards their research project and thesis.


Not sure how to edit, but I was overly harsh: my point about the European programs is that many combine their short duration with requiring a project proposal going in, which often is used by PIs to work on the next ~deterministic step of their existing research program. For students in that situation, a postdoc (or darwinian trial by fire as a professor) is necessary to learn how to lead research.

The above is not always true, but when it is, I view the US system as fulfilling its mission better. Whether everyone is suited to the US system and leading innovative projects is another story.


At least here in Belgium, project proposals (which are indeed required) can come from either the student or a principal investigator. Many of the most talented students pick their own topic, and the topic is accepted or rejected by an independent jury, not by the person who would advise you.


In the US, there's generally a proposal process two years in to determine this, and even there' it's not surprising to see a pivot a year or so after that once preliminary data is in. By pre-agreeing to a problem (and part of the process), too much of the research definition process is skipped.

While half the PhD is about learning how to get to the solution, the other half is in learning how to pick the right problem. In terms of HN, making a startup solve a niche is great, but knowing how to grow it into a bigger space is better.


In Australia it's officially 3 years (with an optional 6 month extension), but out of all other students I knew in grad school there was only one that finished in this time (I'm not sure she ever left her office in that time). I took 7 years, which was not uncommon in my department.

What usually happens is that after the three years (+ extension if you're lucky) is up, you don't have a scholarship any more, and for most people this means they have to find a job to support themselves, which leaves little to no time to work on the thesis. I was fortunate enough to get a teaching position at the same university, but even that was very demanding and my progress slowed to a crawl, though I did eventually manage to finish.

Universities only get paid if the student completes, so they tend to be pretty flexible. If they kick you out at the three year mark they lose; if there's indications you're likely to complete at some point, they have a chance of getting paid, so there's some flexibility here.

I came to the conclusion that ultimately the timeframe doesn't matter - what counts is that you get it done; it takes as long as it takes. A friend of mine was at it for 20 years, but eventually finished. University administration were not exactly pleased with this timeframe and she went through a lot of difficulties because of this, but a remark she made once really stuck to me: "You succeed in spite of the system, not because of it".

Caveat: Both of us are Australian citizens; I can see how the limit would be extremely problematic for international students due to visa rules.


> I came to the conclusion that ultimately the timeframe doesn't matter - what counts is that you get it done

That seems like a bizarre bit of reasoning. When I go grocery shopping, the important part is that it gets done, but that doesn't mean I want to be stuck in the store for five years. At a certain point the cost/benefit ratio becomes insane, not just from a financial point of view but academically as well.


It sounds bizarre as a starting point, but several years in having still not completed, I decided to adopt this attitude as a way of avoiding the immense stress about the thing taking so long. Forgetting about the timeframe and focusing on doing whatever it took (and persisting as long as necessary) is what got me through. Completing was important enough to me that I wasn't going to give up.


i am in the US but have looked abroad for a ph.d. program. there's one in germany that is in english. it is meant to attract students across the world and not just the US. US students are actually rare in the program. but back to my original point, the program has no teaching duties, tutoring or anything of the sort. the coursework requirements are very straightforward and not bloated. it has a very strict timeline, but that's great. you spend 3-4 four years doing nothing but your work, with both an advisor and a supervisor separate from your advisor. it's a great system. i've applied before but went to a US school. i ended dropping out of that ph.d. program with my master's. i sort of wish i hadn't have been in a relationship at the time of going to graduate school as i might have done the german program.

in graduate school, i learned a lot...in the first year. after the fire hose of the first year, i was rather tired of courses, but still had to take 3 full-time courses while teaching and tutoring. ignoring the fact that there were almost no advisors to choose from (that you'd want as an advisor), i had no time to pursue anything being bogged down with homework and exams, which became very tedious. any time i was trying to read on my own, i couldn't concentrate because i knew i had homework and exams to study for.

i went to graduate school based upon a very enjoyable NSF REU. that was fantastic because you picked a problem at the beginning and spent the whole summer working on that as your project. it ended with writing a paper and most of us presented at conferences later in the year. it was an absolutely fantastic experience that had no parallel when i was in my ph.d. program.

and my "failure" in graduate school was even from someone who had never failed on anything. it was like hitting quicksand. i had even read up a lot on how to be successful in graduate school, but it didn't matter. i left because i was miserable and earning $15,000 a year. a mere 6 months or so after leaving i had a job paying $70,000. it would have been sooner, but i was an idiot and thought i'd go back and finish a second bachelor's since i had taken so many courses as an undergrad.


the strict timeline can be a lie. yeah, on paper, you have a 3 year plan, but it fails most of the time so students wind up taking unpaid years to finish.


I did my PhD in The Netherlands. In my field (computational linguistics), the majority of people finished within the allotted four years. The working conditions are good: PhDs are generally employed as regular employees with vacation money and 13th month. The teaching duties were very light (typically TAing and later teaching one 7-week course), giving you a lot of time to focus on research.


Same in France: all the PhD students in computer science I know have finished in 3 years, or 3 years and 6 months (you can request a 6 month extension). You can't carry on after that duration because you have to be funded and be full-time (so no side job).


FWIW, these (IMHO absurdly) short PhD programs basically mean you need to do a postdoc for your CV to even be competitive with US CS PhDs -- assuming you want to do research in a top institution or research lab. So it's far from an obvious optimization.


That was not my experience. I did a PhD in quantum physics in the UK (Imperial College) in 3 years + 6 month extension. I collaborated with Oxford University and I was able to publish in Nature Physics, PRL, etc ... (in other words, the top journals in the field). I was offered post-doc positions but eventually went to industry.

I really enjoyed my PhD as a time to learn and think about deep and difficult questions. Regarding employment, I think it shows that you can commit to a problem. It shows perseverance and drive. And those are good skill to have at any technical job.


My comments should be interpreted as applying to PhD studies in Computer Science.


The view from Europe is that since students in US PhD programs spend their first two years doing courses they come out with similar actual research experience.


The view is wrong. In computer science -- at least at top programs -- students begin research immediately.


In my area, no-one has failed their 3 year plan yet, maybe up to 4 years taken - our area has very little Australian citizens so the incentives are large.

(there's an obvious Soviet joke here about 3 year plans)


Some European PhDs are worthless. Dr. med and Dr. jur (medicine and law) are a joke in Germany. You can obtain a Dr. med after one year of injecting $random_substance into mice and repeating the "experiment" until the p-values spit out by Excel are sufficient for the "thesis" to pass.

PhDs in e.g. Physics are harder though.


Dr.med. and Dr.jur. are not the equivalents of PhD/DPhil, they are the equivalents of MD and JD degrees, which in terms of research output are also a joke in anglo-saxon countries. In the US, medical doctors who wants to pursue medical research usually do a PhD _on top_ of their MD degree.


Oxford also puts PhD students on a pretty strict timeline. I don't think an Oxford PhD is a joke.


Grad school isn't easy for most people. I had quite a few years of rejections and then bam, bam, bam ... three first author acceptances in a row at top conferences. Even Jean mentioned she submitted to several top-tier conferences and didn't get any of those papers accepted.

I'm sorry to hear about your family situation. Internships provide much needed $$$ to supplement the crappy stipend if you are in North America. I lectured part-time and being a TA is also an option. The teaching was one of the better parts of my grad experience. One mistake I made was not work as a free-lance dev while in grad school. I know people who did and it pays well and also looks good on the resume (also a distraction so you maintain sanity).

Be proactive about getting a job after school. Sad to say .. the primary goal during the PhD is to get out of grad school (ideally with a PhD). The old style industrial labs are in bad shape IMHO. Try your very best to get a prof job ... I went for labs but friends who went the prof route got tenure around years 4-6. I am year 8 and still have to deal with work insecurities. The good news is that non-research jobs are plentiful and the PhD gives you a leg up. My overall suggestion ... lower your expectations and find happiness. A CS PhD doesn't mean success or status these days (not sure if it ever did). Good luck!


All I can say is, I was exactly where you are now. You sound like you're on the right track, sometimes getting publications accepted takes a long time. Good luck to you.


For those interested about income during PhD http://www.phdstipends.com/results gives a good overview of PhD stipends offered by most Universities. A grad student at MIT CS doing a summer internship with Google etc. should typically gross at least 54,000$.

Another point that is often overlooked is that if you want a good team at Google / Apple such as working on autonomous car or deep learning etc. you have better chance going if you have a PhD in similar area. I agree that a PhD in Theory/PL is much more a life of mind rather than Machine Learning/Vision/Systems which can be justified as a bet on possibility of a future payout. I know several students at Cornell who joined Google/FB after BS only to be placed in not very interesting teams and later decided to join the PhD program. I think if played correctly a PhD from a good CS school to an extent can help you avoid some of risks faced by middle-aged programmers. Especially those who do not wish to transition into management roles.

Personally I am doing a PhD (Information Science / large scale data mining for healthcare), since I had guaranteed funding along with access to a large medical dataset that would have been impossible to get without being in academia.


> I will say that graduate school was one of the best periods of my life.

Results not typical (https://news.ycombinator.com/item?id=8720640). It's surprising for me to read the author's account and not find even the smallest tidbit of the adversity and nonsensical friction that plagued my friends and I when we were in grad school.

At least the author is at least partially aware of why they had a good experience:

> I am privileged to have had a relatively supportive environment and few additional pressures. My experience is most likely to generalize for other computer science PhD students at top schools, where the funding situation tends to be good and advisors tend to give students a fair amount of freedom.


I bet there's a lot of correlation. I also did my Ph.D. at MIT, and had a similar experience. My first year was horrible and I was probably one emotional breakdown away from therapy, but it was impostor-syndrome stress, not any silliness of the program itself. The rest was fantastic, and changed me and my life for the better.

I was lucky to attend the wedding of one of my former Ph.D. students a month or so ago. At the (not all that large) wedding were about a dozen of his friends from grad school. It was a fantastic reunion, and evidence to me about the truth of what Jean wrote of finding some of the most intellectually fulfilling friendships of her life.

None of this is intended to deny that others have bad experiences, at the same or other schools -- two people having a great experience is just that: two people's experience. I have a good friend whose horrible experience as a biology Ph.D. at Berkeley led her to reject the entire <expletive deleted> field.

There's a lot of material out there that's negative about the Ph.D. process, and it's useful to hear the positive experiences side, which is (IMO) a bit less represented.


Keep in mind that this is what a PhD is like at elite universities. If you take away interesting researches, great interns from big companies, and awesome colleagues provided by those universities, then you may see a different picture of what a PhD life is like.


Pretty much. I got my PhD from the same school and department as the author. The only thing I would say though is it's more like "this is what a PhD is like at a good research group/department". There are a few excellent researchers at "non-elite" schools where I have friends doing very well, publishing in top journals, and being extremely productive. There are also people at "elite" schools doing not so well because their research direction just isn't that school's forte or they didn't encounter the right group of people for their own passions.

In any case, when I applied to grad school, my strategy wasn't really to apply to safety schools after the competitive schools. Instead I applied simultaneously to competitive schools as well as several companies, and decided that if I didn't get into a PhD program at a place I liked, I was going to go to industry or industry research instead.


A few words of caution from a recent Ph.D. graduate: a big problem you will face if you do this is the all-or-nothing nature of the degree. If you are five years in, you will put up with a huge amount of abuse to avoid throwing away that time. Departments and your advisor know this - they will abuse you in lots of ways (stiffing you on reimbursements, making you teach constantly, etc.) because you are stuck - if you leave, you will be perceived as having failed and derive little career benefit from those years of your life. I would also caution that the author's good outcome (assistant professorship at a top tier school, straight out of grad school), would be considered impossibly unlikely in my field, chemistry. I can't speak to CS though.


I quit recently after 5 years. I don't think I regret it - the improvement in my mental health has been dramatic. I found it difficult to trust any of the advice I was getting from supervisors etc. because our interests weren't aligned.

I can say that I'm significantly more highly skilled than when I started though. Knowing that I didn't want to stay in academia, I spent a lot of time working on my transferable skills (i.e. tech).


Sounds like you played it smart developing industry-relevant skills. I lucked out with that myself and got a thesis project that helped me get a software engineering job afterward. This is really one of the things they should emphasize to new students: the odds you will get a good academic gig, even if grad school works out well for you, are very low, so make alternate plans. Glad to hear things have gotten better for you after getting out - good luck!


Right. It's a lot like deciding to set out on a risky mountaineering expedition where success is determined only by reaching the summit. Along the way, there are numerous pitfalls that can be your undoing regardless of your level of preparation. However, adversarial, abusive, non-cooperative overlords can make it arbitrarily hard, and even impossible, for you to succeed. They get to define what "summit" means and they can move the goalposts at a moment's notice. There is no recourse for you, and there is no oversight.

Playing the game without having benevolent overlords is a fool's errand, but many people do not learn this until it is too late.

One of the most important lessons that can be taken away from grad school [1]: "If you compete, never leave your fate in the hands of the judges. In regular life, never leave your fate in the hands of other people."

[1] https://www.t-nation.com/powerful-words/4-things-you-can-lea...


I am doing thesis based masters (kind of micro or nano PhD) in CS at a Canadian university. My experience is same. My supervisor is text-book narcissist, pathological liar, and abusive. He would always hire international students from third-world countries to ensure that he can get away with abuse since he knows they will do anything to get a degree. And the worst part is the whole academic system is designed around exploiting and abusing (maybe not in some top schools) grad students. In may respect, modern academia is new form of slavery.


I have nothing but respect for people who have the grit and skill to go through a PhD program. Another side of the coin - dropping out of PhD studies after 6 months 10 years ago was retrospectively the best thing I ever did. Quitting is fine if one sees better opportunities elsewhere - NOT quitting when the only reason not to quit is an irrational fear of "being a failure" is equal to a sunk cost fallacy. I was quite familiar with the research setting at that point and moving to other things allowed me more space to grow professionally and meet new people with different views of life. Yeah, I choose OpenGL over density functional theory. This was a much better match for my limited math skills (compared to other physicists) and enthusiasim towards shiny things. I would have not quitted unless the new position had seemed like a very unique position, though - which it was.


DFT in particular is a trap - the field is simply not respected by other physicists for a variety of reasons too complicated to get into here. I dabbled in it at first but was able to switch to hard condensed matter because I managed to pull off passing a QFT course in my first year of grad school which combined with solving some experimentalist's problems got the requisite professors to at least listen to me. A lot of DFT people end up as soft money research assistant professors or switched to staff support scientist positions (if they don't drop out and become programmers) because the perception is they never came up with any ideas of their own - they just executed other people's codes. Leave the field to the J-1s.


Eh? This isn't my experience of condensed matter DFT at all. It's a pretty hot area in the UK/Europe at least, and there are lots of postdocs being advertised, as well as a reasonable number of new faculty being hired in the area. Maybe if you're just hitting 'go' on someone else's code, sure, but developing high throughput analysis or higher levels of theory... I mean, until recently, the head of the Cambridge CMP theory group was a DFT guy.


UK and Europe are different. Italy has very active electronic structure groups and UK has people like the group at Cambridge and Rex Godby or Matthias Scheffler at FHI. The funding structure here in the US simply does not reward higher levels of theory, and high throughput analysis is being driven here in the US by Scheffler's acolytes (and Gerbrand Ceder, etc). No less than Peter Littlewood told people here at a DFT conference a few years back bitching about funding "Go to chemistry and materials science departments instead, they will love you"


I agree with your analysis, I did notice that not many groups in the US were doing what looked like interesting new methods/theory, more 'cashing in' on existing code in a fairly short term way. There are people like Louie and Vanderbilt, though. I did my DFT methods PhD in a materials department group, full of Italian/German/English physicists, anyway :)


Vanderbilt and Louie are the real deal, but what they do is much closer to hard CMT. They have students who are real superstars.

As an aside, Vanderbilt was at the same conference. He told me how the same trend chasing and flight to chemistry and materials science departments had taken place among DFT people with high-Tc in the eighties that was taking place with topological materials now, and that this was a symptom of "(long-time sociological) problems I can't fix"


In contrast to sentiment of many replies here, I have really enjoyed my PhD (admittedly I have not yet submitted - I plan to in the summer). I worked for a year prior to starting, and while that was fun, it felt very limited by comparison. Being around smart people who are interested in a variety of topics, and have the freedom to pursue them, makes for an amazing environment. I've learned a huge amount, and I think the tradeoff of low pay for a few years was easily worth it.


The whole experience of doing a Ph.D in computer science is so different to doing one in the experimental sciences. My Ph.D was a constant struggle against nature and her desire to keep her secrets. As a general rule you could expect 98% of all your experiments to fail and usually for totally boring reasons like your reagents had died. Getting anywhere was a real struggle.

I loved writing up my papers as this was at least something I had some control over. I do agree that doing a Ph.D is an experience like no other and one I don't regret.


Looks like what a typical run through Graduate School looks like, but while it seems in the US things are more dragged down it also seems like the opportunities to connect to industry and work in immediate problems is greater

Also this is priceless http://haskellryangosling.tumblr.com/ (done by the blog author)


Do it to max out the education attribute on your char build. Plus, school is cool.

In all seriousness, it's a great time to learn how to think and get comfortable thinking. That said, there are likely many other ways to achieve this.


This is a nice write up. The only thing I found a bit distressing is the "sense of academic achievement" or how one would call it. In my idealized world (à la Hamming's You and Your Research), a great scholar would feel great from doing great science and having a real impact:

* An applied scientist would feel great from solving someone's important problem -- i.e. saving lives, increasing revenue and/or decreasing costs at the end of the day.

* A theoretician would feel great from solving a recognized important problem.

On one hand, the author writes enthusiastically about her side-projects / internships / hobbies. On the other hand, almost all academic remarks end with:

"Our paper got rejected / accepted at [top venue]. We felt sad / relieved."

Without any further remarks about that work -- i.e. it's all "publish or perish" no matter what.

In that way, it sounds as if one was doing research for the sake of producing papers rather than solving important problems / contributing to our civilization by increasing its stock of knowledge. Perhaps that is the sad state of the current academia.


TLDR: I found a difference between these two senses of academic achievement when looking back at oneself's work:

1. (Idealized) "I/we did a great job solving that important problem."

2. (Current Academia?) "Phew! I/we finally scored another paper at [top venue]." ("And I can now graduate / get a tenure / [something academic career related].")


Are PhDs expensive, do you get paid? In Europe you are actually a full time worker for the university and get paid a living wage (bit lower than industry but it ramps up I guess).


In Computer Science, PhD students at any reasonable institution are invariably funded (stipend + tuition & some fees). Most who are in non-theoretical fields get research assistantships, which pay for them to do their PhD without any teaching responsibilities. The pay is in no way comparable to industrial salaries (OK, it's comparable; it's WAY less). The University of Massachusetts Amherst, where I am a faculty member, pays our PhD students more than $2,500 per month, plus benefits. This is a pretty competitive figure. However, many PhD students do summer research internships at places like Microsoft Research, IBM Research, Facebook, and Google, and these pay pretty handsomely.


Furthermore the student stipend may go further because many needs at a big university may subsidized. These include student residences, gyms, health care, etc. To maintain the same lifestyle after leaving the ubiversity was more expensive.


I should have clarified: the above is for US institutions (and I believe Canada). I do not know how things are in other countries.


This is not true in the UK. You are paid a tax free stipend (currently ~£14k) and not considered a full worker - you certainly don't have the rights that full time staff have.

There is no salary progression, bonus or pension and your funding cuts off at 3/3.5/4 years. After that, you aren't eligible for any govt. benefits as you are a full time student.

In some parts of the UK £14k can go a long way, but UK PhD students definitely aren't equal to full time staff as it stands.

The pay can be higher with industry partnerships (£30k tax free!!) but this is the more usual research council level.


You might get a tax free stipend... :) Depends on whether you get a scholarship! (I have a PhD student of mine who has a "fees only" scholarship - she is employed doing other things at the university but that's separate from the PhD.


Thankfully there has been a shift recently by the funding bodies to increase the stipend beyond the required minimum. Many in biomedical programmes are offering £18-20k a year.


Not true for Portugal too. For my PhD I got a scholarship of 980€/month (too much below standard market rates for software engineers around here) and was not considered full time worker, so I had no benefits, no insurance, nada, zip.

Although I consider it to be one of the best times that I had, in professional terms I didn't got much from it unfortunately..


They pay a subsistence living.


$21k USD in the USA, I've often been envious of my European colleagues!


I'm genuinely interested: jeeves weighs in at ~20k loc. at 50 lines of code/day that's a 1 year endeavor. I know there's a lot of research upstream, TAing, writing papers, and so on, but 50 loc/day is something very achievable while doing lots of other things. My naive question is why/how does it seem like the final result is way smaller than what I would expect for 7 years of research?


I am not sure what you are getting at. Are you counting productivity by LOC?

To sum-up: - This is not her only project, just the main project - She has taken grad courses, doing a grad course in MIT will easily be more work than average developer do in whole year. - She had to study several research areas before choosing her main project. It is unlike I read the tutorial of currently trending distributed system Apache Spark and now I know everything about state of the art research. It is going through dozens of research papers in that domain, understanding all the concepts, identifying the gap in existing research, figuring out how to fill that gap. - You will fail several times before finding a good approach. The author have mentioned she had tried with Scala first, later switched to Python. - Even the results are good, you have to write the paper which is not trivial, and getting it accepted before getting rejected several times. - Worst case scenarios are also likely. Mid-way in the project you have found out someone else has recently published a paper. Now your work is worth-less because it is already published. Or you have found it there is an obscure paper that also outlined the approach you have taken, now you have lost the novelty factor of your paper.

TL;DR: 200 lines of some UI JS code is not same as 200 lines of new deep learning algorithm in TensorFlow.


With normal programming (e.g Python web development) you have a really good idea of what you are building before hand. That's not necessarily the case here. Also it sounds like they did various rewrites at different points.


Since having a PhD is not a prerequisite for most jobs, it only makes sense if you are genuinely interested in science. The system is geared to support this.

Unfortunately, the system is also biased for the young and single and child-free. When I tried entering the PhD program, I was ready for significant income drop (I had been a successful software engineer for 10 years before). But I also had a 10-months daughter (she is almost 2 now). And this proved to be incompatible with the requirements of the job. I am not a father who neglects his family needs and focuses on the job. So I moved back to consulting, working from home most of the time.

I don't see an adequate solution for this. Of course, young and single will be preferred to survive the grind, and forcing the employers to do otherwise is unfair to them. Perhaps there is something else I can do to make a difference.


Ouch. That sounds very mundane. I hope this person finds something fulfilling in the future.


[flagged]


> Fuck this guy

Personal attacks are not allowed here. (You're factually mistaken, also.)

More generally, please don't just vent. Instead, share your experience. Your comment does both, but the balance is wrong. If it had more of the latter, I think the community would find it interesting and probably respond with good discussion.


[flagged]


>his

try again


PhDs seem like a journey to nowhere. What's the point? I guess I just don't see what you get out of it that's valuable enough to devote 4 years.


There are some areas for which a PhD is a pre-requisite. I currently work at a synchrotron. Accelerator operators, engineers etc. do not all have PhDs, but when you get to the science side I think every single member of the science team has a PhD.

A PhD is the most straightforward path towards involvement in science that needs big, expensive, sophisticated equipment.


The point is to train you to be a scholar, someone who can do original research in the field, as is expected of university professors. To that end, a PhD program includes coursework advanced enough to bring you up to the current state of the research in the subject, and a period of research apprenticeship with your advisor, where you begin to contribute to the field yourself. That's valuable enough to devote four years to if scholar is the job you want. I wouldn't recommend it for any other purpose. Way too much pain.


Lots of people say they got something out of a PhD. Even if they are all biased, doing a PhD is a culturally acceptable way of being unemployed and goofing off.

http://lemire.me/blog/2015/02/25/hopeless-ones/


You are not unemployed when doing a PhD, nor are you goofing off. Not sure if your comment was supposed to be sarcastic somehow, but you are very misinformed.


I would say you are underemployed.


You are underpaid, but I would not say underemployed. The benefit society gains from a PhD is high, mainly because of the training students get from doing a PhD. But there is an over-supply of students who want to do PhDs as opposed to the budget for science, so the pay is very low.


I think lots of people get the idea of PhD is worthless learning from PhD graduate from some mediocre varsities. Getting PhD from any elite school (like MIT in this case) or elite research group is absolutely priceless.


well i wanted to be a professor so getting a phd seemed like the most direct route.


de Broglie's are rare.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: