Monday, November 15, 2010

Why I'm leaving Harvard

The word is out that I have decided to resign my tenured faculty job at Harvard to remain at Google. Obviously this will be a big change in my career, and one that I have spent a tremendous amount of time mulling over the last few months.

Rather than let rumors spread about the reasons for my move, I think I should be pretty direct in explaining my thinking here.

I should say first of all that I'm not leaving because of any problems with Harvard. On the contrary, I love Harvard, and will miss it a lot. The computer science faculty are absolutely top-notch, and the students are the best a professor could ever hope to work with. It is a fantastic environment, very supportive, and full of great people. They were crazy enough to give me tenure, and I feel no small pang of guilt for leaving now. I joined Harvard because it offered the opportunity to make a big impact on a great department at an important school, and I have no regrets about my decision to go there eight years ago. But my own priorities in life have changed, and I feel that it's time to move on.

There is one simple reason that I'm leaving academia: I simply love work I'm doing at Google. I get to hack all day, working on problems that are orders of magnitude larger and more interesting than I can work on at any university. That is really hard to beat, and is worth more to me than having "Prof." in front of my name, or a big office, or even permanent employment. In many ways, working at Google is realizing the dream I've had of building big systems my entire career.

As I've blogged about before, being a professor is not the job I thought it would be. There's a lot of overhead involved, and (at least for me) getting funding is a lot harder than it should be. Also, it's increasingly hard to do "big systems" work in an academic setting. Arguably the problems in industry are so much larger than what most academics can tackle. It would be nice if that would change, but you know the saying -- if you can't beat 'em, join 'em.

The cynical view is that as an academic systems researcher, the very best possible outcome for your research is that someone at Google or Microsoft or Facebook reads one of your papers, gets inspired by it, and implements something like it internally. Chances are they will have to change your idea drastically to get it to actually work, and you'll never hear about it. And of course the amount of overhead and red tape (grant proposals, teaching, committee work, etc.) you have to do apart from the interesting technical work severely limits your ability to actually get to that point. At Google, I have a much more direct route from idea to execution to impact. I can just sit down and write the code and deploy the system, on more machines than I will ever have access to at a university. I personally find this far more satisfying than the elaborate academic process.

Of course, academic research is incredibly important, and forms the basis for much of what happens in industry. The question for me is simply which side of the innovation pipeline I want to work on. Academics have a lot of freedom, but this comes at the cost of high overhead and a longer path from idea to application. I really admire the academics who have had major impact outside of the ivory tower, like David Patterson at Berkeley. I also admire the professors who flourish in an academic setting, writing books, giving talks, mentoring students, sitting on government advisory boards, all that. I never found most of those things very satisfying, and all of that extra work only takes away from time spent building systems, which is what I really want to be doing.

We'll be moving to Seattle in the spring, where Google has a sizable office. (Why Seattle and not California? Mainly my wife also has a great job lined up there, but Seattle's also a lot more affordable, and we can live in the city without a long commute to work.) I'm really excited about the move and the new opportunities. At the same time I'm sad about leaving my colleagues and family at Harvard. I owe them so much for their support and encouragement over the years. Hopefully they can understand my reasons for leaving and that this is the hardest thing I've ever had to do.

81 comments:

  1. congrats on the move. the academy's loss, google's gain. looking forward to not hearing about what you are up to (per google policy of course :)

    -Remzi

    ReplyDelete
  2. It more or less means that we will not be able to benefit from your active participation in the Sensys community. I am happy for you but sad about this great loss to the academia and the sensor networks community? I do not think that Google will let you do sensornets.

    ReplyDelete
  3. Phoey! I'm someone who hopes to go back to school for an academic degree in the hopes of eventually working in industry. If the best academics feel they are disconnected from what Google and other such companies are doing, I question if getting an academic degree really puts me in the running for such a job someday.

    Best of luck to you, Matt. :)

    ReplyDelete
  4. Grad Students: One more top 20 tenure track job just opened up!

    Matt didn't talk about the money, but I assume it is somewhat better than in academia (but only by 30%-50%)

    Giving up tenure is a sacrifice, but probably less so for someone with strong applied interests and who enjoys so much hacking on industry scale problems.

    Good luck Matt and we hope to hear more from you.

    ReplyDelete
  5. Matt: your paragraph that started out "The cynical view..." ended up as a personal statement. Freudian slip? :-) It's been a long week of finalizing the decision I'm sure.

    We'll miss your papers and your students. But looking forward to collaborating with you in industry!

    Given your readership, I feel obliged to reiterate that there are lots of ways to have impact, and doing University research -- with all its responsibilities -- is definitely one of them. And one of the most fun and direct, in many ways. No boss, no short-term bottom line, no annual ranking and rating nonsense, you can even skip the meetings to do your work. The best people and the best ideas wander the halls at the good schools, you're constantly bombarded with smart energetic particles. It's a powerful and exciting place to sit. And it stays that way year after year.

    Industrial development is also a great place to have impact on the physical world and the world of ideas. Lots of ways to make a difference, and we all need to keep collaborating.

    I know you'll have a blast at Google and I hope the change will do you good. Enjoy it. Improve the world from that end. Keep in touch. Keep sharing your ideas.

    Joe Hellerstein

    ReplyDelete
  6. So you believe that the real impact of CS comes from industry, not academic. And trying to build a "big system" in university is somehow less meaningful because they can not beat industry. of course you can get ideas and experiences by doing so, but why don't you directly join the company like Google. what is your suggestion for PhD in system field? Do they need to go into real industry and solve the real problem rather than thinking themselves in campus and build a far less impact prototype?

    ReplyDelete
  7. Thanks for the comments. Joe is absolutely right that University research can be a great career and a wonderful way to have impact. I want to emphasize that I don't feel that industry is "better" than academia or vice versa. It just depends on how you want to spend your time. It's like asking whether you want to play baseball or soccer - which is "better"? Depends on what you want to do.

    I absolutely recommend getting a PhD (see my earlier blog post on that) in systems for those that want to have a big impact EITHER in academia or industry. A PhD is not only useful in an academic setting and gives you deep insights and expertise that you would not get going straight into industry. That said a PhD is not for everyone.

    ReplyDelete
  8. Wow! I'm a grad student in distributed systems right now and you just expressed every feeling I have for why I want to work in industry at a place like Google.

    Maybe one day we'll get to interact :)

    ReplyDelete
  9. Congratulations! I hope you'll still blog about your work, even if in a very indirect way. It's been a pleasure reading your blog and getting insights about the academic world.

    ReplyDelete
  10. Perhaps for systems research, that is true. It is more about engineering, to make stuff work practically, at a large (but not over large) scale. For other areas of CS, it might be different.

    Good luck Matt!

    ReplyDelete
  11. Good luck Matt--I do hope you will still be publishing some of your innovative ideas so all can share and benefit too :-).

    Your decision raised two points which I would like to share, both motivated, as in the case of Joey, to your general audience. (1) I am distressed to read and hear about another stellar top researcher having trouble getting funding. We as a community should be more open to good and often risky ideas, especially in academia, and should encourage rather than suppress them "till they reach an acceptable stage of maturity" (as is often said!). This often squelches innovative ideas, and as in your case, might encourage researches to search elsewhere, in this case Google. (2) Again, aimed especially at young PhDs reading this blog, I owuld like to point out that one of the most rewarding aspects of academic research is the continious mentorship of young talents. I find the teaching/mentoring of students as one of the most attractive, and maybe the most rewarding aspect of the academic research process.

    best of luck!

    ReplyDelete
  12. Good luck Matt! The only way to up deploying sensors on volcano-tops was to work on the clouds. Have fun.

    ReplyDelete
  13. Congrats, Matt. Sorry to hear you are leaving Boston, but that place on the other coast is not too bad either.

    ReplyDelete
  14. Welcome to Seattle! Stuart and I used to miss many restaurants in Harvard square but Seattle has amazing salmon selection that is unbeatable. =)

    ReplyDelete
  15. Dear Matt,

    Are you not afraid that you won't be as challenged in industry as you have been in academia? Is the glory of professorship not really worth it (especially given the high quality of life at Google)? Or perhaps the notion of academic glory, as seen through the eyes of a soon-to-graduate PhD student, is an illusion?

    Pragmatically speaking, I think there is much to be said for having a high-quality lifestyle at Google and doing other fun stuff on nights and weekends (such as being a good father!). Would it be incorrect to view your decision as that of father vs. professor?

    I, for one, welcome our new professor-stealing overlords.

    ReplyDelete
  16. money money money money money money money money money money

    justify it all you want, it's mostly for money. i like money too. no problems here, but dishonest not to mention it.

    ReplyDelete
  17. How about your blog? Dose Google have restrictions about the personal blog of its employee? I hope you can share your experiences in Google about "big system", especially comparing with academia. that will help us to figure out whether the problem tackled in campus is the "real problem". thanks in advance.

    ReplyDelete
  18. Tomasz - In academia, you have to come up with the challenges yourself and then spend a lot of time convincing the rest of the world that the challenges are interesting. In industry, the challenges are coming at you from the outside world. There are plenty of ex-profs at places like Google and Microsoft, so there is strong evidence that there are enough intellectual challenges there to keep someone like me occupied.

    Anon re: money. Harvard pays its faculty (especially tenured faculty) quite well. Money motivates people in different ways and I think bringing that up only clouds the issue. Yes, I get paid more at Google; no, that is not why I am leaving. Would I stop working altogether if I made a zillion dollars in the lottery? No way. I want to do work that is interesting and be happy doing it.

    Webcraft - Google lets me blog as long as I'm not revealing company secrets or saying nasty things about the competition.

    ReplyDelete
  19. I don't know you but this article does stop me in my tracks, because the future of the web is unpredictable. I learn that you're joining Google on the day FB.Com opens...and a veritable war between the two is in full swing. Too bad that they couldn't give you a leave of absence at Harvard. But we don't know your deepest personal motivations. Were they once academic and now entrepreneurial? If so,your decision makes sense.

    ReplyDelete
  20. he's already on a leave of absence.

    ReplyDelete
  21. This comment has been removed by the author.

    ReplyDelete
  22. Don't go. We'll miss you a lot.

    ReplyDelete
  23. If this is the hardest thing you've ever had to do, you have a pretty fantastic life.

    ReplyDelete
  24. Seattle++;

    We're lucky to have you in the area!

    ReplyDelete
  25. Good luck with everything down at Google! Sounds like you'll be having a great time hacking away in Seattle.

    Sadly, this means my graduating class and I won't be having you for CS61 or CS161 during our years here -- from what I hear, you were a fantastic professor in many respects. Harvard will miss you! Many of my upperclassmen (your past students) have already mentioned so.

    ReplyDelete
  26. Where in Seattle are you moving to?

    Right in the rainy season too! Also, which Google office will you be at?

    Seattle News

    ReplyDelete
  27. "saying nasty things about the competition. "
    Why? Saying nasty things about themselves is one thing, but saying nasty things about competitors is a different thing altogether.

    ReplyDelete
  28. Nice to know Google joins the ranks of MSR and Yahoo Research in the fine art of 'poaching' top professors :)
    Good luck!

    ReplyDelete
  29. I bet after looking at Google's patent portfolio, Matt realizes he is five years behind, and therefore, feels redundant if he stays in academia. :)

    ReplyDelete
  30. Matt: It is a hard decision and one that young faculty members often have to make for the opportunities against its costs. For sure, it is a golden age of the computer science industry, just as Bell Labs provided a compelling intellectual environment a generation earlier, and perhaps not so much for the science and technology funding in a "zero base" budgeting environment.

    While there are many reasons for staying in academia versus industry and it ultimately comes down to personal preferences, I have to split with you on the notion of "making an impact".

    Where else do you find fresh batches of talent year in and year out to inspire and make them think in ways that few else can? But then, you surely know that for having taught, if not inspired, the competition to your new employer!

    ReplyDelete
  31. @Yuhong Bao. Libel lawsuits. These things are better handled by marketing departments.

    ReplyDelete
  32. I have to say, it would be amusing if you eventually end up in Facebook, don't you agree?

    ReplyDelete
  33. An older related blog post by Amin Vahdat

    http://idleprocess.wordpress.com/2010/06/15/doing-big-science-in-academia/

    ReplyDelete
  34. Very interesting that it can also go the other way. I worked for a while in industry (not google though) and even though I had great colleagues, great liberty, and a lot of technical fun, I became bored after a few years. I'm now doing physics related fundamental research and have a faculty position at another university and I would never trade in science for any other job ever again :)

    Anyway, good luck, hope it works out well!

    ReplyDelete
  35. Good luck Matt! We will welcome you over at UW anytime. I agree with what my good friend Prof. Hellerstein had to say as well as the comments of others. The key is to figure out what you most enjoy doing. I see my impact as much larger on students (mainly the undergrads!) I enjoy working with my grad students and seeing them develop the most.

    But, I ask that folks (especially at the NSF and on NSF panels) to wake up. People like Matt and the work he does are what we should be funding. Stop the risk adverse incrementalism that is taking over academic computer science! This is the tre tragedy I see in Matt's leaving Harvard.

    ReplyDelete
  36. Great loss to CS Academics (in the short term) and great gain for Google. Any graduate student can only dream to work on projects like RoboBees.

    ReplyDelete
  37. Best of much Matt - time to hack some andriod code!

    ReplyDelete
  38. I totally agree with what you tried to say with 'scale'. At company like Google, you could have thousands of cores on hand easily and run your MPI program with real data anytime without waiting in the queue for a long time.

    ReplyDelete
  39. I hear Amin Vahdat is joining Google too.

    ReplyDelete
  40. Matt: I understand your post and decision completely...and am incredibly (and selfishly) saddened by it at the same time :(

    On the bright side (...well, realistically cloudy side), you will love Seattle and Google and I wish you and your family the best!!

    ReplyDelete
  41. Joe H. has clearly drunk some serious Kool-Aid. Having been in both academia and industry, I can assure you that annual performance reviews are nothing compared to the stress of conference program committees controlling your career as is the case in academia. And "The best people and the best ideas wander the halls [and], you're constantly bombarded with smart energetic particles" described Google, Microsoft Research, etc., just as well as it does academia.

    It's definitely different folks for different strokes, but I think folks like Joe are actually doing students a huge disservice by exaggerating the upsides of academia and minimizing the quality of the environment and work at top-notch industry places like the ones I mentioned. It's a great way to create more professor-clones, but not fair or realistic to the students.

    ReplyDelete
  42. @Anon-5:12pm: "...but I think folks like Joe are actually doing students a huge disservice by exaggerating the upsides of academia and minimizing the quality of the environment and work at top-notch industry places..."

    To be fair to Joe H., if you read his comment carefully, he only described the upsides of academia, and never minimized the quality of the environment and work at top-notch industry places.

    ReplyDelete
  43. I can definitely understand Matt's decision. As an academic researcher working on problems in the area of social media, I am envious of the vast amounts of data available to researchers working at Google, Microsoft, Yahoo, and Facebook, and their ease of access to the social graph and user interaction data.

    ReplyDelete
  44. Good luck! After several years in full-time industry, I was thrilled to escape to academia. I do think that Google will be a great place for years to come, but I wouldn't trade my teaching job for a Google job, even at double the salary. On the other hand, I've never had to think much about raising money---I've mostly had more than I could spend---and when I was in industry, I noticed to my great surprise that I really missed being around students.

    As I have gotten older, my priorities have changed. In my early 30s, it was all about great code. Now that I'm in my late 40s, it's all about great students. As you said, soccer and baseball. Anyway, I wish you great success in your new sport, and I hope you will travel back East to tell us about it in person.

    ReplyDelete
  45. I don't know why systems-researchers in academia even believe(d) that they were solving anything interesting.

    In a way - what's done in academia as systems research can at best be called 'trivial in every respect'. What's done in industry (as part of systems research) can at best be described as "large scale and cumbersome but intellectually trivial".

    If what people in theory do can be termed research, then it is simply ridiculous to even use the same term for what's done in the area of systems/networking.

    I remember this incident while attending 2001 SOSP: before the start of one of the sessions, the session chair ( I remember everything/everybody vividly, but not taking any names here) mentioned that one of his colleagues told him that SOSP is a brain-dead conference. The audience laughed - probably in disbelief. Having attended a few SOSPs, and having followed most of their papers for many years, it seems to me - nothing could be closer to the truth :)

    ReplyDelete
  46. Bhaskar KrishnamachariNovember 17, 2010 at 10:47 AM

    I wish you all the very best, Matt. I'm sure you will continue to grow and meet your potential at Google.

    Your technical and service contributions to the sensor network community have been outstanding, and I'm sure you'll bring the same passion to your work and community in the future.

    I liked your comment about soccer and baseball. Everyone should do what moves them most deeply. I have always most enjoyed the process of learning and sharing what I learn with others. I really love working with students, and helping them grow. I could not imagine a different life than being an academic. But for someone like you whose main motivation is to build and deploy systems that a lot of people use, this move to industry makes a lot of sense.

    ReplyDelete
  47. w.r.t. the question about Industry vs. Academia, I think that Research labs, such as IBM Research, still offer the best of all worlds - Real "physical" impact, together with Academic focus and opportunity (even requirement) to publish academic papers.

    ReplyDelete
  48. Matt -- wish you all the best at Google. I'm sure that you will have fun building great stuff there.

    At the end of your post one June 6th, you said, "I'll post a more personalized account of what it took to navigate Harvard's tenure process". Even though you'll no longer be in academic, we would be interested in your account of the tenure process.

    ReplyDelete
  49. @annonymous..."...Research Lab such as IBM Research still offers best of... "
    Bell Labs may be the last of the Industry Research Lab that brought so many disruptive technologies to the market. May be XEROX PARC was up there with innovative ideas. IMHO, Industries Research investment has been going down.

    ReplyDelete
  50. Shit! I was looking forward to do my MS and PhD under you at Harvard. Looks like I have to change my plans!

    But Matt you go ahead..........make Google proud, not to mention fill your pockets as well.

    ReplyDelete
  51. Since my name was mentioned and this discussion touches on topics where I spent more than 30 years talking and thinking about, I feel I could add some more perspective.

    If past is prologue, impact will come both from academia and industry. The National Research Council published two reports asking where did the new multibillion dollar industries come from, and they listed 19 examples in the report "Innovation in Information Technology" in 2003. Examples include much of the technology that you use everyday: The Internet, Relational Database, RAID Storage, Graphical User Interfaces, RISC computers, Speech Recognition, .. .

    In every case, there was a back and forth between ideas advanced in industry and in academia.

    How have things changed in the last decade or so?

    Industrial research is arguably worse off: while Microsoft created a new large research arm and places like Amazon, Google, ... have exciting jobs for systems people with PhDs, Bell Labs is gone, Xerox PARC is not what it was in the 1970s, and many long time research labs have had their research horizon shortened.

    For academia, the cutback by DARPA in systems research since 2000 was a heavy blow, which increased the competition for NSF funds, which made funding much harder for everyone. The new administration and new DARPA leadership have returned to valuing academic research. DARPA funding in academic IT is now on an uptick. However, it takes a while for this surge in funding to percolate through researchers and reduce the demand on NSF.

    In terms of being able to do research with impact, open source software and cloud computing are boons to academia. We can do research with access to the source code of industrial quality software and work at a scale that's makes it possible to evaluate your research ideas believably. Systems research in academia would be a lot more challenging without these two innovations.

    Possibly more of the interesting systems problems in this decade are so hard that they need to be tackled by larger groups of people than the traditional single professor with his or her graduate students. Industry regularly works pretty well regularly in larger teams.

    However, systems at Berkeley, for example, has a long tradition of working in groups of 3 to 6 faculty and dozens of graduate students to take on projects with potentially high impact. (I'm wrapping up my 10th such project so that we can start the 11th!)

    Even if working in larger research groups is more important than in the future than in has been in the past, teams can be successful in academia as well as in industry.

    I don't think most systems people are going to industry for the money nor are those going to academia so that they can be called Professor. We're fortunate to have chosen a field were there are great jobs in industry as well as academia. I'd like to believe systems people are largely following their hearts, with the former being pulled more by the chance to directly shape products than millions of people can use and the latter pulled more by the chance to shape the lives of the next generation of brilliant young computer scientists.

    ReplyDelete
  52. Good lord Matt, I thought you would have gotten into Google a long time ago :D


    I do not think that Google will let you do sensornets.

    A certain former professor once said that sensornetwork research is like digging a hole, jumping into the hole, then building a ladder to climb out of it. He was very recently working on sensornets.


    The best people and the best ideas wander the halls at the good schools, you're constantly bombarded with smart energetic particles.

    Before I came into industry, I thought the same. After some exposure I realized that's not really the case: there are very many good people in the industry too. Think about it: we begin with a group of people who are very smart. Some have the opportunity and patience to go through the PhD program. Others, for various reasons, decide not to. What's the ratio like between the two?

    The very good people I know may not have had the training to write eloquent introductions, but they are nonetheless equally sharp and analytical. They may not be able to add a 'Dr' or 'Prof' in front of their names, but they can be just as good at building systems as any professor, usually even better (because of their experience). When these folks get good ideas, they don't seek personal glory by publishing papers or blowing their own trumpets. Because (1) from real-world experience others know how good they are and (2) they're typically working for the good of their organization and teammates, not for themselves. Those who work for themselves only in the industry will gradually become ostracized by others, it's very obvious.

    I think folks like Joe are actually doing students a huge disservice by exaggerating the upsides of academia and minimizing the quality of the environment and work at top-notch industry places like the ones I mentioned. It's a great way to create more professor-clones, but not fair or realistic to the students.

    Sad but true, it's self-perpetuating. The worth of a professorship is derived from how other folks look up to them: "Look here, I'm famous, I write lots of papers, I have many students, therefore you should fund me!" How many things are wrong with that statement?

    I can definitely understand Matt's decision. As an academic researcher working on problems in the area of social media, I am envious of the vast amounts of data available to researchers working at Google, Microsoft, Yahoo, and Facebook, and their ease of access to the social graph and user interaction data.

    There are many different aspects of building things that work: Design (architecture) -> Implementation -> Deployment -> Maintenance -> Performance measurement -> (repeat). I know quite a number of people who stick with performance measurement, which, while still important, is simpler (or at least less tedious) than the other phases. The worst kind of researchers are those who can publish simply because they have access to data that others don't have. If you are one of those, please figure out a way to anonymize the data etc. and make it publicly available.

    In a way - what's done in academia as systems research can at best be called 'trivial in every respect'. What's done in industry (as part of systems research) can at best be described as "large scale and cumbersome but intellectually trivial".

    Say I'm writing an academic paper. It's probably enough to show that something works for 95% of the time. In the industry, because of the scale and the need for things to really work, 95% is not good enough. And in some cases, the solution to make the remaining 5% work can be really intellectually stimulating and challenging, but it'll be treated as inconsequential by the academic community. The academic reviewers can really be brain-dead, looking purely at the numerical results. So perhaps you're right after all :D

    ReplyDelete
  53. Good luck at Google, it's great to be in a place where your unique talents are best leveraged. It used to be that the smartest people in the world could exercise their gifts best in universities by producing high impact research and high quality teaching. These days, however, many faculty spend the bulk of their time doing grant-writing, administrative work, producing reports, handling committees, writing departmental newsletters, tracking budgets, handling expenses, etc. A lot of this work could be offloaded to administrative staff who don't have PhDs but they are all unionized and busy playing tetris.

    ReplyDelete
  54. Things aren't that much worse in academia than they were 30 years ago, and things aren't that much better in industry now relative to academia than they were 30 years ago. There are pluses and minuses in both career choices now just as there were then.

    I've never heard anyone say that all the smart people are in academia. I don't know anyone who believes that.

    Its just that academia does have SOME of the smart people, which is why the history is that big innovations usually have both an academic pieces and an industrial piece.

    Academia also has really tremendous but inexperienced young people lining up trying to come to join your group. Unlike, say, history, information technology is a young person's field, so this is a nice feature of academia.

    (By the way, I'd believe more in your arguments if you believed them strongly enough to be willing to put your name on your opinions.)

    ReplyDelete
  55. I don't see a single comment that says "this is shocking", "I'm disappointed", "why would anyone do that?" and yet when talking to people offline I do hear reactions like these.

    I'm posting as a cowardly "anonymous" person because I'm coming on the job market this year and don't think can speak my mind freely otherwise.

    Here are a few points:

    - I refuse to believe that 5 months at Google was what it took to convince you that you like the life of a hacker more. It's much more plausible that in the 8 years of the grind (running for tenure) you discovered that this life is not for me. The post could have been much more interesting if you could talk more about this aspect. More importantly, why leave now when the worst part is over? (the last one is the most puzzling question actually)

    - Is this the best use of your skills? In the hierarchy of things, being a nameless hacker in a cubicle is generally at the bottom and being a free-thinking "visionary" scientist is at the top. If I got tenure at Harvard and wanted to have more impact in real life, I'd think about doing a startup or getting a leg into IT policy making, or picking up a cause (say something like one-laptop-per-child) and pushing it forward; you know unlimited possibilities? Now, what's the difference between you writing that code and a student of yours from your OS class writing it? Berkeley, Harvard, and the society at large invested in you over the past 13+ years and maybe society will benefit more if you made the best use of your skills, instead of doing things that others can potentially do as well.

    - What does this say about the Harvard tenure process? Were they not able to detect that this person prefers to hack instead of being a Professor?

    - What about your students? They worked hard and played a role in getting you tenure. You are going to leave them hanging in there, just like that?

    My questions will come across as overly harsh, so I'm going to end the comment by talking about some positive things:

    - I don't think money has a role to play in this. You won't be earning 1.5x more, certainly not 2x more - more like 1.3x or something. That is nothing for someone who is in the 150k+ bracket anyway (my estimates might be off, but you get the point). So it's certainly not about the money.

    - You have a point about the fuzzy line between research and industry when it comes to systems people. Researchers don't have the resources to replicate the problems or study real properties of systems, how can they even begin to solve anything.

    - I'm sure that you are more intelligent than me and maybe you see a bigger opportunity that I can't wrap my head around and some years from now, you can point back and say - "look idiots, I made the right choice".

    ReplyDelete
  56. "Cowardly" anon - You make some very good points, and believe me, all of these things have been going through my mind for the last several months as I contemplated this decision. It was not easy and who knows if it will ultimately be the right decision. I do feel that after 7+ years as a faculty member and 5 years of grad school that I need a change of pace.

    As for leaving now - I should have left Harvard long ago, but it was not until I got to Google that I realized I could be happy in this kind of environment. I'm not sure tenure changes anything. I don't envy the tenured faculty at Harvard who seem to do even more travel and sitting on committees than junior faculty do. (Those who do that seem to love it, but my point is that I'm not sure the worst actually *is* over for me.)

    I really like your vision for pushing things forward, and under other circumstances I think I could have that kind of impact in academia. There are a lot of reasons why (a) I think Harvard is the wrong place for that, and (b) I am not the right person to be doing that kind of thing. I am not David Patterson or Nicholas Negroponte, much as I'd like to be. I'm personally so much happier as a builder than a manager. Perhaps that means I will not accomplish as much in my career; so be it. If I'm happy, and get to spend time with my family, and don't spend every day of my life grunting and complaining and feeling like I should be working harder, that is much more important to me than getting my picture on the cover of Wired.

    I've thought many times about doing a startup. I don't think I have the constitution for it. I should have done it when I was 25, didn't have a family to support and could throw my whole life into it. The great thing about being at Google is that I can focus my time on precisely the technical stuff that really interests me and not have to worry about raising VC and getting customers and all that business stuff that I know nothing about. I admire people who can do all of that.

    I am not the only tenured faculty member to give up the academic life for Google or other companies -- all of the other ex-faculty I've spoken with at Google cite exactly the same reasons for leaving -- they want to have more impact, spend more time doing interesting technical work and less time doing academic bullshit.

    I can't speak to the Harvard tenure process, but let me suggest that had they not tenured me, they would have looked like idiots and wouldn't be able to hire anyone else good in systems for a long time.

    Most of my students are pretty much set and several are about to graduate. I have three students who will need to find other advisors, though for the most part they are early enough in their careers that they can readily switch. I am going to continue working with them and funding them until this happens - I am not too worried about it, though I feel badly for having made this change.

    Actually the difference in compensation is substantially greater than you estimate. As I mentioned before, this was not my reason for leaving.

    And thank you for the kind words, but I can't be that smart if it took me 8 years in a faculty job to figure this out.

    ReplyDelete
  57. Cowardly Anon again.

    @Matt: Thanks for the quick response. Your reasons make more sense now. A few points:

    - I do buy the argument for doing startups early in life when you don't have wife and kids (especially kids). I can totally imagine you wanting to spend more time with your son and not having to worry about doing a startup that runs out of VC money and puts his future education at stake. But at the same time the lifetime job guarantee coming with a tenure is the safety net for doing startups as well i.e., if it fails, you always have your job. Also, these days startups are fighting over tech talent like hungry wolves. At Harvard you're sitting in the talent pond and can groom them as you please (talking about undergrads, who can then be the fuel for your startup).

    - About you not being Negroponte or Patterson. Don't think when they were your age, they knew about the impact they'd end up having. For all I know, you are one of the more influential thought leaders in the (young) system's community right now. This blog itself is evidence of that. People do listen to what you have to say; most of the time this is all you need for having impact.

    I think I got my answers. It's good to know that you wrestled with the same questions for a long time before coming to a conclusion. Good luck!

    ReplyDelete
  58. (By the way, I'd believe more in your arguments if you believed them strongly enough to be willing to put your name on your opinions.)

    This is fantastic, that's exactly what I feel when I read reviews! Since our thoughts are aligned, would you consider getting all the CS conferences' reviewers to reveal their names? I think that would surely help reduce some of the ridiculous feedback we get!

    Thanks for all your help Dave!! But sadly, before then, I have to continue being an anonymous coward :(

    ReplyDelete
  59. There are obvious reasons why you need blind reviewers for papers so that people can give their honest technical opinion. When I serve on program committees, I go out of my way to ensure that reviews are civil.

    Not obvious to me why its OK in our field to express strong public opinions on non-technical issues without having the courage to give your name. There were nasty things published about Matt, Google, academic research, ... there were only because they could say it without anyone finding out who said it.

    The way I was raised, if you don't have the courage to make your statement to someone's face, then you should keep your opinions to yourself.

    Do you really want to sustain the "weenie" stereotype of people in IT?

    ReplyDelete
  60. For a while I've been tempted to disable anonymous comments on this blog, but in the end I think it makes things more interesting. I reserve the right to delete comments (anonymous or not) that I feel are uncivil :-)

    ReplyDelete
  61. @Anonymous said... 30% ...

    Not to mention the 10% pay rise coming Google employees way in the New Year:)

    ReplyDelete
  62. This comment has been removed by the author.

    ReplyDelete
  63. Google seems to be swallowing up a lot of faculty nowadays. http://bit.ly/aOncMH

    Microsoft used to do so, but don't think Apple or IBM really made a habit of it.

    ReplyDelete
  64. The accusation that professors propagate the 'academia > industry' myth is unfair, since I'm sure most students get to intern in industry (research/dev) several times during their PhD to see the 'real world' for themselves, before deciding where to go.

    What does appear to be true is that increasingly, more academics seem to be interested (at least partly) in 'immediate-impact' industry-ish research, which maybe due to various pressures (visibility, funding, tenure, etc).

    If one finds herself sitting in academia, impatiently bemoaning that her research is not having 'immediate-impact', she probably has the wrong job.

    Ideally, an academic would have far greater freedom to do 'pure, exploratory research', but unfortunately, the 'culture' built up over the years makes this increasingly hard. It is very tough to secure funding or publish papers if the research is not necessarily 'immediately valuable'.

    ReplyDelete
  65. @Joel, I made a comment about Bell Labs might be the last Industry Research Lab to influence so much technology - theory to applications.
    If I am correct, Dr. Paul Horn retired from IBM Research and at NYU providing leadership for collaboration between industry and academic research? IBM Research was big in inventing DRAM etc. (personal computer era?) It seems like Google Research in distributed computing era has potential to become what Bell Labs was in communication era. However most computing research shall come from Universities and academics?

    ReplyDelete
  66. hope you like the rain. it doesn't stop except for a few days in the summer in seattle!

    ReplyDelete
  67. Well the good thing about industry is that everyone is treated with respect, even the guy who is put on a PIP (Performance Improvement Program - normally at the end of which 90% of them are fired, I think in places like Google/Microsoft this is something like 5% of the total population per year or something like that). So it does not matter whether you are yourself a Professor/Phd/___(fill in the blank here), you will "have to be civil" to everyone, otherwise you will have your back exposed for everyone to see (if you know what I mean!). However, in industry you will also have to put up with a lot more of the usual "corporate PR BS" by rolling over your eyes as and when necessary.

    Academia on the other hand is totally different, as there is an asymmetrical relationship between a Prof and his grad student - (or should I say Master/Slave?). Academia definitely seems to have a higher percentage of a**holes (or it atleast appears to be, as academia provides more freedom for people to be at their absolute best/worst without the necessary requirement of having to maintain the minimum level of decency)- and whenever I think of academia one of the immediate images that come to my mind is one that of - "Professor Brian S. Smith" from PHD comics.

    The other best/worst thing about academia is the tenure process. I was recently reading some discussion on the P != NP problem in light of recent failed attempt made by Vinay Deolalikar - and someone made a very interesting comment saying that although we don't yet know whether P is or/is not equal to NP, but we can confidently predict that this question is very unlikely to be solved by a tenure chasing academic!

    ReplyDelete
  68. I look forward to the blog post titled: "What I do in the '20 percent time'" :)

    ReplyDelete
  69. I find the superficial level of this discussion somewhat disheartening. Why no mention of public good? The point of academic work is general advancement and dissemination of human knowledge. By working for the private sector, you relinquish your ability to share the fruits of your intellectual work with the rest of us, at least for as long as the fruits are ripe (giving your private employer a competitive edge). The important question is why is academia not able to offer you the tools / money / infrastructure you need to achieve your research goals while the private sector can. The simple answer is that corporations reap substantial benefits from academic research but do not share them back with the public which has funded the research (they invest in basic research but it's a very small fraction of the benefits). It's a fundamental imbalance which companies such as Google are not in a hurry to correct.

    ReplyDelete
  70. Anon Re: "superficial level". It's true that working for Google means that I can't freely share my work with the world, at least not directly. I can of course publish papers on my work at Google, so that isn't going to change. Google does release some things as open source, but that's a tiny fraction of the work they produce.

    Still, I think you are overstating the public good here. The primary benefactors of my work at Harvard were my students, of which I had roughly 80 (or so) undergrads a year pass through my intro systems class and another dozen or so grad students in my grad seminar. The number of people who read my papers is vanishingly small, on the order of a few hundred, and those that could actually put the ideas to any use even smaller. At Google, the work I am doing is impacting the Internet experience for millions of people every day. In what way does this not constitute public good?

    The real question we should be asking is why universities expect professors to work themselves to the bone and spend so much time on things other than research, which is ostensibly of the greatest importance to society according to your definition. If universities were really in the game to promote society they would do a much better job at protecting (especially junior) faculty from administrative overhead and having to work 70+ hours a week just to keep on top of things. A place like Harvard has billions of dollars at its disposal but can't seem to figure out how to hire adequate administrative staff and teaching assistants to assist faculty in carrying out their mission. Talk about imbalance!

    ReplyDelete
  71. Matt, Do you think "publish or perish" is hurting academics environment? When you said why universities expect professors to work so hard and spend much time on things other than research, do you also mean teaching?
    I wish to go back to state school for graduate degree and that's why I am wondering if teaching and interacting-guiding students will become less important than Research at most reputed schools?

    ReplyDelete
  72. Maulik - I don't include teaching as part of overhead. Teaching is at least as important as research in an academic setting, and is obviously a substantial portion of what faculty are paid to do. I personally love teaching and got a lot out of it, and for the most part never felt that teaching negatively impacted my research productivity (at least relative to all the other faculty out there!). On the other hand, I do think that universities can do more to support junior faculty and reduce teaching burden through more staff support - but Harvard recently cut the number of teaching assistants assigned to courses, citing cost concerns.

    All that said, as far as tenure and promotion are concerned, I have yet to find a good university that gives a shit about teaching. As long as you are cranking out papers and doing great research, nobody cares if you are a brilliant teacher and mentor or not. Everyone says that teaching matters but think of it this way: none of the external promotion letters will have anything to say about your teaching, since outside people won't know what you are up to in that regard. The lesson is that you are not rewarded for being innovative or taking risks when it comes to teaching, so just teach the same class for 8 years without changing the syllabus and focus on writing more papers instead. The students will suffer but at least you will get tenure.

    ReplyDelete
  73. Maulik - lest you get the wrong idea here, I think Harvard is about as good as any university gets in terms of faculty support. (Though Harvard has more money than most of them.) My comment is directed towards universities in general, which expect faculty to juggle a tremendous number of responsibilities - research, teaching, mentorship, internal and external service, fundraising, hiring, outreach. My experience so far at Google has given me a much better appreciation for the model of "if people are working too hard, maybe it's time to hire more people." Of course, Harvard has a relatively small CS department, so I don't know if my perception is skewed by that.

    ReplyDelete
  74. Matt, thank you for your candid reply and it didn't give any wrong idea. It's rather ubiquitous at most universities. Most professors are professional (and remember being graduate student :-) ) who put students needs reasonable priority and their passion and involvement for research brings better teaching. Challenge again is balance.
    Top universities have advantage not only with bigger budget and endowment but also bigger pool of relatively higher quality graduate students to pick for Teaching Fellows. That makes big difference.
    Google and Facebook are top of the game and only hire very best so can afford to hire more greater minds but other relatively more mature Tech companies have gone through relentless cost cutting and lay offs in last decades and situation may not be the same.

    ReplyDelete
  75. I think it would be interesting to come up with an alternative to the current system. Here's an (extremely) ridiculous alternative, meant to bring smiles (hopefully):

    We get rid of the tenure system. Professors are appointed for (say) 5-year periods, selection of the next batch of professors are performed by the incumbents, based on a wide pool (other universities, industry, pseudo-industry-academic-whatchumightcallit-labs). We get rid of the 'top-conference' paper selection process. Instead, papers are accepted as long as they are scientifically correct. The worth of a paper (and author) is judged based on its impact in the real-world. For instance, we can say, "Wow, this-or-that organization does this fantastic thing used by millions of people, and Dr. XYZ is the brains behind it!".

    By eliminating 'iron-rice bowls', we don't get 'master-slave' faculty-student relations. By getting rid of the tenure process, spanking new PhDs get to focus on long-term research issues rather than getting papers and tenure. By getting PhDs to cycle between industry and academia, we get faculties who are in sync with industry, teach more up-to-date material to their students, and (hopefully) can garner industry funding support during their 5-year tenures at their universities.

    Yes I know, haha very funny :)

    ReplyDelete
  76. I had the pleasure of meeting you and hearing your lectures while taking your class at Harvard, and although I can't say that I relished learning about zombies and children and lock and keys, it was great to meet a professor who I could tell really loved the subject he was teaching.

    As a student, I could also always tell that you were meant for so much more than academia, and although I too love Harvard, I know that you will blaze another great trail at Google. Good luck and well wishes :)!

    ReplyDelete
  77. Matt,

    You were one of my favorite editors during my PhD days (you were an editor for my TOSN paper). Sorry to see you leave. But glad you figured out what you want in life.

    ReplyDelete
  78. GangstaRapperLA ATTTT GMAIL DOTTTT COMMM

    Dave

    Whassup Holmes !
    I wish I could roll wid you. I mean Computer Archtecture came naturally to me. I shudda stuck wid da plan and pursued a graduate program at Wisconsin or wid you.

    Them head hunters at Intel wouldn't leave my trail. I got hired in for a CAD/Design position and the promotions kept coming. A brother needs his bling, ya know what I am sayin?

    I hear you bench press, how much do you bench?

    ReplyDelete
  79. Thank you...

    I'm currently in a two-year postdoc (humanities) and am considering not pursuing an academic career anymore (desire to re-engage with broad learning, disillusion with academia and its culture, personal circumstances, etc.)

    As you know, this is a complicated decision-making process fraught with clashing emotions. Yet I find much comfort and a guiding compass in the stories of other academics who left the industry.

    Again, thanks...

    ReplyDelete
  80. Well, I am a tenured CS faculty at a ranked place, and I would say the following: Since before I joined Academia, I have felt that getting a tenure is perhaps tantamount to becoming a multi-millionaire (since, both give an option to lead a almost-retired life). However, after getting tenured, I see the dangers too. One can become too complacent. So, I am also considering to do something about it -- not as drastic as what you have done (bravo!), but maybe take a 2-3 years leave (thanks to the tenured life!) and indulge in something more "dynamic".

    Irrespective of the above, I have started to feel (after tenure) that being a professor for the rest of my life will certainly not leave me satisfied (even if I consider the best case scenario of having accomplished the most as a professor). Somehow, writing papers, getting grants, graduating students, etc. will not have satisfied me at the end of it all. Impact? I doubt we as *individual* professor make much of an impact (maybe, as a community we do) -- or maybe, some do, but even that level of "impact" somehow wouldn't leave me satisfied at the end of it all.

    ReplyDelete

Startup Life: Three Months In

I've posted a story to Medium on what it's been like to work at a startup, after years at Google. Check it out here.