5 ms·

Why I'm leaving Harvard (for Google)

127 points by leibniz 16y ago

jacobolus 16y ago
Big loss for Harvard. Matt Welsh is a fantastic teacher.
- snowmaker 16y ago
  Agreed, I had him as well and he was a big inspiration. Perhaps the beginning of a broader shift in industry vs. academia?
  - fragmede 16y ago
    Or rather, a shift back to industry funded research a-la Xerox Parc and Bell Labs, among others.
  - Lewisham 16y ago
    I think the pull is that Google is a very academic place (based on visits and talking to people who work there), and if Google's interests align with yours, you're basically doing the research you would have already done PLUS you get paid better PLUS you don't have to fight for funding on an almost daily basis PLUS you actually have data. Google is a fabulous, exciting, wonderland place for lots of software engineering professors/students. The question is whether Google is a career. Places like Google, Microsoft Research (and, perhaps Facebook) allow for the freedom that academics crave, but that's only three places. If research-style engineering goes out of fashion in favor of enterprise middle-management hell, you're going to be left high and dry if you don't have a publishing record to fall back on when you try going back to the Ivory Tower. Fortunately, those places have enough cachet that you can say "I worked at Google" and no publishing record for those years are forgiven, but if you don't have one (say, you went half way through your PhD), coming back is going to be hard. That's why the decision is so difficult: you're taking a bet on whether the benefits taken for granted in academia are going to continue in industry. A broader shift would have to show more companies than those three, in order to provide (for want of a better word) "safety" that your worklife goals aren't in jeopardy.
    - moultano 16y ago
      Thankfully for their job security though, research projects at Google regularly launch as parts of major products.
    - kingkilr 16y ago
      They're not new though, just the next generation. The previous one had places like IBM and AT&T employing researchers in this manner.
  - kenjackson 16y ago
    That shift happened back in the .com boom of the late 90s. It has never really reverted back. It's still the case that top-flight PhDs are going to work as developers at companies (versus academic or industry researchers). I think it is a testament to how interesting the work in our field is.
- swies 16y ago
  Yep, definitely my favorite professor. Too bad to see him go, but it sounds like Google's a great fit for what he wants to do.
teichman 16y ago
Matt: what will happen to your graduate students?
- seiji 16y ago
  My name is Matt, so I feel qualified to reply. My first joke thought was: "I bet Google will be happy to hire them too." A moment later I realized it's probably true.
  - Estragon 16y ago
    You probably would have gotten positive points for this if you'd left off the extra "My name is Matt" joke.
gojomo 16y ago
A previous hint of the pull of Google for Welch was: http://matt-welsh.blogspot.com/2010/10/computing-at-scale-or-how-google-has.html http://matt-welsh.blogspot.com/2010/10/computing-at-scale-or...
klochner 16y ago
Someone has to say it: So he can leave google for facebook
anrope 16y ago
"...overhead and red tape (grant proposals, teaching, committee work, etc.)..." I feel like a professor shouldn't consider teaching "overhead". If you're teaching classes, you should be putting at least as much effort into research as teaching. Prof != post-doc. If your passion is in research, then a move to industry (or post-doc) sounds like a good choice.
- rguzman 16y ago
  I agree that teaching should not be perceived as "overhead". However, the best place to do research is within academia (large research universities), where there is a very complete and strong infrastructure to do research. Yes, the system is broken in many ways, but it is still the best. People who really have a strong passion for teaching and not as much interest in research go teach at universities that specialize in teaching more than in research e.g. Reed College. Last, at a sufficiently advanced level teaching and research are indistinguishable.
- Smerity 16y ago
  He may be complaining about all the overhead that goes with teaching rather than the teaching itself. I'm tutoring this semester for a Data Structures course and a painfully large portion of the time is spent dealing with things that really aren't teaching. Trying to deal with plagiarism (which even goes as far as assignments posted on RentACoder), using ancient marking systems (WebCT is well and truly evil), handling students who obviously don't care at all but feel entitled and so on.
  - anrope 16y ago
    That's fair. Still, in my experience (which is nothing special, 4 years undergrad, in second year of masters) there are definitely professors who consider teaching an unfortunate side-requirement of their research. And I don't think it's unique to my school. I think this sucks. I have been working as a teacher's assistant for a few semesters now, and I hear what you're saying about students who really don't care. I think especially in engineering, you get students who chose the major just because they got good grades in math and science and heard engineering pays well. Then those students complain when you ask them to work hard (or even just work). Sidenote: WebCT :( haha. My school paid (and will continue to pay) a ridiculous amount for some enterprise course scheduling/billing/etc. software that would have been a great project in software engineering (e.g. large project management). All while we're having major budget issues.
mbyrne 16y ago
Short version: Why I'm leaving Harvard (for Google) is because I want to be an engineer and not an academic.
- jlees 16y ago
  s/engineer/researcher/
moultano 16y ago
He's describing something that was the biggest factor for me not deciding to go back to school. I came to Google and started working on search engine spam. I also periodically read papers on "Adversarial Information Retrieval" as it is called in the literature. Only a handful of the papers I read had useful ideas, and I don't think there's anything that can really be done to improve the situation. There are certain things you can't really research without access to the real thing. Information Retrieval is one of them, and I'd be equally skeptical of parallel computing research unless the university has a 1000 node cluster to work with. This sucks for the world, because the research in these things that Google does doesn't really escape to advance the state of the art elsewhere.
- kermit_de_fro 16y ago
  Minor picky point, but universities have access to such computing power via time-sharing at national labs if it is not available from their own department.
  - moultano 16y ago
    And I do think that a lot of good parallel computing research gets done. It's just a lot harder.
- Lewisham 16y ago
  Sadly, most academics realize that they're data-starved, but industry isn't able to/doesn't want to work with academia to help gain access to that data. Even if you do get access, you're NDA'd so hard that publishing useful papers is a challenge. It's a shame, because universities are essentially free labor, just on much longer timeframes and lower probability of immediate success. I understand why things are the way they are, but I'd love to find some way of reconciling this issue. It's already killed a research project of mine before it started.
  - moultano 16y ago
    Google realizes this too, and there are a lot of people here that would like to help. There are a few huge issues that always come up. 1. Paranoia about giving competitors the edge they need. 2. Paranoia about giving spammers the edge they need. 3. Privacy. It's nigh impossible to anonymize user data enough to release and still have it in a useful form. The AOL logs debacle hurt everyone. One of my friends had a really interesting take on this. His belief essentially was that the industry advances when people switch companies and (legal or not) take their institutional knowledge with them. Unfortunately, academia is hard to transfer into.
svrocks 16y ago
Matt made assembly language and multithreading entertaining to learn and accessible for undergrads, which was no small feat. So those of you bashing his teaching need to step off

127 points by leibniz 16y ago

jacobolus 16y ago

Big loss for Harvard. Matt Welsh is a fantastic teacher.

snowmaker 16y ago
Agreed, I had him as well and he was a big inspiration. Perhaps the beginning of a broader shift in industry vs. academia?
- fragmede 16y ago
  Or rather, a shift back to industry funded research a-la Xerox Parc and Bell Labs, among others.
- Lewisham 16y ago
  I think the pull is that Google is a very academic place (based on visits and talking to people who work there), and if Google's interests align with yours, you're basically doing the research you would have already done PLUS you get paid better PLUS you don't have to fight for funding on an almost daily basis PLUS you actually have data. Google is a fabulous, exciting, wonderland place for lots of software engineering professors/students. The question is whether Google is a career. Places like Google, Microsoft Research (and, perhaps Facebook) allow for the freedom that academics crave, but that's only three places. If research-style engineering goes out of fashion in favor of enterprise middle-management hell, you're going to be left high and dry if you don't have a publishing record to fall back on when you try going back to the Ivory Tower. Fortunately, those places have enough cachet that you can say "I worked at Google" and no publishing record for those years are forgiven, but if you don't have one (say, you went half way through your PhD), coming back is going to be hard. That's why the decision is so difficult: you're taking a bet on whether the benefits taken for granted in academia are going to continue in industry. A broader shift would have to show more companies than those three, in order to provide (for want of a better word) "safety" that your worklife goals aren't in jeopardy.
  - moultano 16y ago
    Thankfully for their job security though, research projects at Google regularly launch as parts of major products.
  - kingkilr 16y ago
    They're not new though, just the next generation. The previous one had places like IBM and AT&T employing researchers in this manner.
- kenjackson 16y ago
  That shift happened back in the .com boom of the late 90s. It has never really reverted back. It's still the case that top-flight PhDs are going to work as developers at companies (versus academic or industry researchers). I think it is a testament to how interesting the work in our field is.
swies 16y ago
Yep, definitely my favorite professor. Too bad to see him go, but it sounds like Google's a great fit for what he wants to do.

teichman 16y ago

Matt: what will happen to your graduate students?

seiji 16y ago
My name is Matt, so I feel qualified to reply. My first joke thought was: "I bet Google will be happy to hire them too." A moment later I realized it's probably true.
- Estragon 16y ago
  You probably would have gotten positive points for this if you'd left off the extra "My name is Matt" joke.

gojomo 16y ago

A previous hint of the pull of Google for Welch was: http://matt-welsh.blogspot.com/2010/10/computing-at-scale-or-how-google-has.html http://matt-welsh.blogspot.com/2010/10/computing-at-scale-or...

klochner 16y ago

Someone has to say it: So he can leave google for facebook

anrope 16y ago

"...overhead and red tape (grant proposals, teaching, committee work, etc.)..." I feel like a professor shouldn't consider teaching "overhead". If you're teaching classes, you should be putting at least as much effort into research as teaching. Prof != post-doc. If your passion is in research, then a move to industry (or post-doc) sounds like a good choice.

rguzman 16y ago
I agree that teaching should not be perceived as "overhead". However, the best place to do research is within academia (large research universities), where there is a very complete and strong infrastructure to do research. Yes, the system is broken in many ways, but it is still the best. People who really have a strong passion for teaching and not as much interest in research go teach at universities that specialize in teaching more than in research e.g. Reed College. Last, at a sufficiently advanced level teaching and research are indistinguishable.
Smerity 16y ago
He may be complaining about all the overhead that goes with teaching rather than the teaching itself. I'm tutoring this semester for a Data Structures course and a painfully large portion of the time is spent dealing with things that really aren't teaching. Trying to deal with plagiarism (which even goes as far as assignments posted on RentACoder), using ancient marking systems (WebCT is well and truly evil), handling students who obviously don't care at all but feel entitled and so on.
- anrope 16y ago
  That's fair. Still, in my experience (which is nothing special, 4 years undergrad, in second year of masters) there are definitely professors who consider teaching an unfortunate side-requirement of their research. And I don't think it's unique to my school. I think this sucks. I have been working as a teacher's assistant for a few semesters now, and I hear what you're saying about students who really don't care. I think especially in engineering, you get students who chose the major just because they got good grades in math and science and heard engineering pays well. Then those students complain when you ask them to work hard (or even just work). Sidenote: WebCT :( haha. My school paid (and will continue to pay) a ridiculous amount for some enterprise course scheduling/billing/etc. software that would have been a great project in software engineering (e.g. large project management). All while we're having major budget issues.

mbyrne 16y ago

Short version: Why I'm leaving Harvard (for Google) is because I want to be an engineer and not an academic.

jlees 16y ago
s/engineer/researcher/

moultano 16y ago

He's describing something that was the biggest factor for me not deciding to go back to school. I came to Google and started working on search engine spam. I also periodically read papers on "Adversarial Information Retrieval" as it is called in the literature. Only a handful of the papers I read had useful ideas, and I don't think there's anything that can really be done to improve the situation. There are certain things you can't really research without access to the real thing. Information Retrieval is one of them, and I'd be equally skeptical of parallel computing research unless the university has a 1000 node cluster to work with. This sucks for the world, because the research in these things that Google does doesn't really escape to advance the state of the art elsewhere.

kermit_de_fro 16y ago
Minor picky point, but universities have access to such computing power via time-sharing at national labs if it is not available from their own department.
- moultano 16y ago
  And I do think that a lot of good parallel computing research gets done. It's just a lot harder.
Lewisham 16y ago
Sadly, most academics realize that they're data-starved, but industry isn't able to/doesn't want to work with academia to help gain access to that data. Even if you do get access, you're NDA'd so hard that publishing useful papers is a challenge. It's a shame, because universities are essentially free labor, just on much longer timeframes and lower probability of immediate success. I understand why things are the way they are, but I'd love to find some way of reconciling this issue. It's already killed a research project of mine before it started.
- moultano 16y ago
  Google realizes this too, and there are a lot of people here that would like to help. There are a few huge issues that always come up. 1. Paranoia about giving competitors the edge they need. 2. Paranoia about giving spammers the edge they need. 3. Privacy. It's nigh impossible to anonymize user data enough to release and still have it in a useful form. The AOL logs debacle hurt everyone. One of my friends had a really interesting take on this. His belief essentially was that the industry advances when people switch companies and (legal or not) take their institutional knowledge with them. Unfortunately, academia is hard to transfer into.

svrocks 16y ago

Matt made assembly language and multithreading entertaining to learn and accessible for undergrads, which was no small feat. So those of you bashing his teaching need to step off