Saturday, March 13, 2010

And More Fun News....

Right after Stuart Shieber sent me news on CS undergrad earnings, Harry Lewis sent me two additional links.  The first is a nice Boston Globe piece on the increase on undergraduate majors in the sciences at Harvard.  

The second is much more amusing.  Apparently, the Harvard Robobees project has made #1 (that's right, we're number 1!) on Sean Hannity's list of the 102 worst ways the government is spending your tax dollars.  Now, I try to stay apolitical on this blog, but I have to say, I'm impressed by Sean Hannity's lack (well, actually, more like a complete absence) of acumen in understanding the nature of scientific research.  A look at the Robobees home page, I would think, would certainly suggest that there's important scientific and engineering questions underlying the long-term challenge of building a robotic bee.  Of course, maybe the Hannity camp just objects to the government spending money on science generally, I don't know.  I'll go on record as suggesting that nobody from the Hannity camp bothered to look at the Robobee home page.      


Harry Lewis said...

Perhaps Hannity thinks it will mean unemployment for US soldiers if the bees replace them on scouting missions into Afghan caves. Does he have a clue why the government thinks this stuff might be worth developing?

Anonymous said...

Harvard != Engineering

Anonymous said...

Professor, I have lots of respect for you and your research. However Hannity has all the rights to raise questions regarding the bees project.
I am also a grad student and I personally know that out of a grant of 1M only 100K goes on buying equipment and the rest goes on the stipend+tuition of grad students, conference trips. So, ultimately the biggest chunk of a funding goes to the private university.
Every american, therefore has a right to raise questions about these big grants, specially if the taxpayers money is funneled to the private universities.
Please note that, I am not against scientific research. I have read the project description of the bees project, and it is a novel and useful project. However 1M is too much!!!

Harry Lewis said...

$1M is cheap for this work, if that is the actual number. I doubt even Hannity would think it could be done for less. I'll bet he thinks it's just a silly project that shouldn't be done at all. Pure cheap shot. If he had realized that it would save soldiers' lives, he wouldn't be making fun of it.

Yes, when you have a research contract like this, you pay grad students in part by sending them to conferences, so they will wind up with a net and not just a fish. It's part of the way we invest in the future economic productivity of society as well as getting a deliverable product. Great deal for the taxpayer -- try contracting with the private sector and see what $1M gets you.

Michael Mitzenmacher said...

Anon #3 and Harry: The project is actually $2 million per year, for (up to) 5 years. This is a huge, interdisciplinary project, supporting a number of professors (something like 10) and a large number of graduate students and postdocs. As Harry says, it seems like it's actually an amazing bargain; a lot of interesting stuff is likely to come out of it.

Anon #3: Hannity certainly has the right as an American to raise questions regarding how the government spends money. That's not what he did. He listed it as #1 on a list of "worst ways the government is spending your tax dollars". I see no actual questions raised; indeed, I see no sign that there's any actual understanding of the project at all. Go to the Robobees home page again and see what the research is actually about, and what the possible applications are. I would then appreciate a reasoned argument why you (or Hannity, or anyone) thinks the price tag is too high.

[I should point out that I am not on the Robobees grant, I just think it's a cool research agenda.]

Anonymous said...

Prof. Lewis and Mitzenmacher, I am not against the robobees project.
As, I wrote before, I think it is pretty useful and lots of interesting stuff will come out of it.
I am only against the spending part of this project. I am not an american citizen and the country where I come from, I am pretty much sure this research will not cost so much.
I am not saying that the researchers are committing fraud or something. I am just saying that the money spent on this project is too much. The government should not spend so much money on the tuition, which infact goes to the Harvard university. Also, I donot know about conferences (i have only published a paper, in a conference in US), but I am sure a conference in Europe costs more than $1500 per person.
We need some austerity measures and this is what I am trying to say!!!!
Again, I am myself a grad student in CS, and I know that robobees is a cool and a useful project..

Anonymous said...

Lol!!!....this is really amusing to me. Here I was planning to write to Matt for a postdoc, since I am so impressed with his recent talk at Sensys, and want to real do some real research. Sean Hannity is a real joker. I do hope the funding for robobees continues, just the power systems work on that project would be enough to keep the project going. Swarm robotic systems havent been built and use of something like Robobees for surveilanc is emmense. I still cant believe that joker like Sean Hannity make more money than us. Boohoo!!!

David Andersen said...


You seem to have the belief that money spent on equipment is better spent than money spent on grad students, postdocs, and faculty.


First of all, that's backwards: The more research you can do for the amount of equipment you have to buy, the better off you are. Computer science is fortunate, not the opposite, that we don't have the costly equipment requirements that a lot of other fields. Even in systems, where we buy a lot of computers, it's not like equipping a solid state matter lab.

Second of all: In research as in many things, you get what you pay for. Ask yourself honestly: Does $70k spent on one of Matt Welsh's students buy you more or less than $70k spent to support two students at the university of nowhere? Three students? How much volume does it take to make up for having the work done by one of the top people in the field? Can you even do so?

You also have a fairly mistaken impression of where money spent on tuition goes. At Carnegie Mellon, for example, a decent portion of the tuition that we pay for graduate students goes towards ensuring that all students in the department are funded. I've seen what can happen at universities when faculty can threaten to withhold funding from their graduate students, and it's not pretty.

Oh yeah, and some of it goes to make sure they have insignificant things like medical coverage, etc. Y'know. Useless waste of money stuff.

And finally - $1500 for a conference, when you send a student to one or two conferences per year? That's only 10% of their stipend cost alone. And the benefits of conferences are large -- you should go read some of the research about the function of, e.g., DARPA, the Israeli military, and other national funding bodies in forcing scientists to talk to each other and share information. The bottom lines are very clear: Information sharing is a huge benefit to the discipline as a whole and it increases the productive output substantially compared to the small amount of money it takes to enable that sharing.

I think you're barking up the wrong tree. A culture of austerity does no good on its own. The question we should ask is: How do we maximize the amount of productive research we do per taxpayer dollar input? (Keeping in mind that there are many outputs of that research, including the training of Ph.D., MS, and BS students, as well as spinoffs, startups, patents and licensing, and publications). It's a much more nuanced question than the one you asked, and in that light, I'd politely suggest that a $2m/year investment in robobees might actually be a very good use of funding.

Anonymous said...

Prof. Andersen,
Since you have replied with such confidence, I ask you to share the breakup of costs of a project where the funding was >=1M USD.
Unless, you corroborate your arguments with solid facts, I am sorry, I cannot trust your "rant".
I said earlier also, that I have nothing against the robobees project, I find it cool and useful. I just think that 2M a year is too much money for this project. Infact, there are so many great projects in CS, where an independent observer, can see that the money spent was too much.

Anonymous said...

And yes, I know that training Phd students is a good use of the grant money.
But, should the universities, not share the costs of training Phds???
Please donot think as a professor, think as a normal american. Definitely spending tax dollars on phds, rather than on military is a good thing. But, is it better than not spending money on health care and public schools???

David Andersen said...
This comment has been removed by the author.
David Andersen said...

That's very easy. The AIP project is funded for about $1.x million, across four faculty members + each of our graduate students, over three years. It's divided equally between faculty. From the budget we submitted (a bit less was funded): My share of it is about $125k per year. Of that, $60k goes graduate student stipends + tuition. $30k of it goes to the university and my department to cover overhead expenses -- everything from power to administrative support to building maint, etc. About $10k pays for faculty member time. $7.5k goes to equipment. About $7k per year goes to hire undergrads for the project. $6k per year goes to travel. My co-PIs budgets are similar-ish.

For a $2M grant, just multiply it up. In particular, if you look at the number of faculty funded by something like an NSF expeditions, it's huge compared to a normal grant. Therefore, it's pretty safe to assume parallel scaling + some amount of money for unification costs (programmers, travel, etc.).

And honestly - the onus is upon you at this point to justify why you think $2m/year is too much for the robobees project, given the potential returns on investment from it.

Should the universities not share the cost of training Ph.D.s? Remember that a university is an organization that turns input money (from students paying tuition and from other organizations sponsoring research) into education and research. While universities typically produce a net gain in value, that value then leaves the university and hangs out gainfully employed or turned into a product at a company.

And re thinking as a normal person: I am. I'm also a U.S. taxpayer, remember, and I want to get the most quality of life/etc., out of my tax-paying investment, just like everybody else does. There are two questions to ask: What should the allocation be between, e.g., the military, research, and health care; and what do we demand of the people to whom we allocate that money. My comment was about the second. The former is not a matter for one person, it's a matter of national policy. As a professor who receives taxpayer funding, my responsibility is to do as much high-quality research & education as I can with the budget I receive.

And no: It's not clear that we should be spending more on health care and less on research. This is not a simple question. What pattern of spending is best for the long-term health, competitiveness, and security of our nation and of the world?

[post revised to correct budget information submitted vs. received]

Anonymous said...

@Prof. Andersen.
Thanks for the detailed comments. I am still not satisfied with your answer. But, I am happy that you replied to my question.
I was not expecting an answer from you.

David Andersen said...


I'll post my last on this (thanks for putting up with my abuse of your blog, Mitz. :).

I observe at this point that you have failed to answer any of the questions that I or other commenters have asked you. Nor have you responded in a substantive way to the further information that people have taken time to provide in response. That comes across more as trolling than as a serious attempt at discourse on what is an important and relevant question. I'm done feeding the trolls. :)

Michael Mitzenmacher said...


I agree with you, in that I don't think anonymous has a serious discourse in mind. Thanks for responding and providing the info.

Anonymous said...

Reading the first 20 entries in Hannity's list it is clear that the man was just looking for projects with silly titles, robobees being one of them. One shouldn't read into it too much. My reaction (comming from a country where anti-intelectual sentiments run very deep and do a lot of damage) is lamenting the popularity and influence of this clown.


Anonymous said...

David A:

Out of 125K, you are saying that 60K goes to the administration (overhead plus half of tuition + stipends). That's a lot!! For heating, electricity and administrative support?

I think the point is that these universities generally have huge endowments, and it seems that when they get such a grant, they take half and at least some of that is put into their savings, and why should tax payer money be put into the savings account of the university?

I don't have a problem with tax payer money funding research at a private university, I just think all of the tax money should go directly into research and not to the university. So if it pays the electricity bill that is racked up during a research project, fine. But it should be accounted for and not just go into the bank.

--A different anonymous.

David Andersen said...

Hi, Anon2 --

I've decided I need to make a blog post of my own about the cost breakdown. It's actually an interesting question.

No, 60K doesn't really go to the administration. Another chunk of the tuition that's paid for grad students actually ends up paying for the faculty again (because grad students take classes...). As an example: I directly pull in research funding to cover three months of my salary. The other 9 months are paid by the university. The money that comes from "the university" to pay me during the academic year... comes from tuition. Which I pay on behalf of my grad students.

It's a bit of a weird accounting mechanism to account for things like grad students taking classes from multiple faculty, taking out-of-department classes, etc. But I'd probably estimate that of the ~$30k I pay in tuition for a student, about ~$15k of that actually goes towards faculty salary that covers both teaching and advising (which == research).

(One way to think about the overhead is to ask how much it costs to rent office space + administrative / secretarial support for one person for one year. A lowball estimate is probably $1-2k/month for highly oversubscribed space -- and we have labs, grant administrators, lawyers, etc.)

I don't know the numbers on the sources of endowment funding, but my understanding is that most of the endowment actually comes not from "profit" on research funding, but from donations and perhaps things like licensing. The margins on sponsored research are pretty slim.

I'll try to write up a more careful analysis of some of this from the perspective of my own funding. It's interesting to try to chase down where the money actually goes.

Noel said...

Anon the different, think about implementing your scheme. It would be horrendous. How much more would it cost to do this than just paying some share of the overheads?

Anonymous said...

"Does $70k spent on one of Matt Welsh's students buy you more or less than $70k spent to support two students at the university of nowhere?"

That's pure arrogance! Good out and smell the roses you dimwit!

Anonymous said...

I am sorry, but I have to defend anonymous here (at least in part).

I am a faculty member in a top 25 CS department. I have been on several large-ish grants, so I know what I am talking about. I won't go so far as to say money is "wasted", but you can be sure that the money is not being spent as an average citizen reading a grant budget might expect.

For starters, we have the significant university overhead. Maybe it is justified, maybe it is not, but the overhead represents a huge percentage of the money being spent by the government.

11 faculty are associated with the robobees project. If you do the numbers, this works out to roughly $200K per faculty member per year. By way of comparison, an NSF CAREER award is about $80K (for one faculty member) per year. $200K is probably enough for 2 months of summer support for each faculty plus 2 grad students (plus equipment, travel, etc.). Are 22 graduate students working full time on this project? Are 11 faculty members spending a full two months of their time (each year) working on the project?

I am not suggesting there is anything wrong with the Robobees project, or even that the money is wasted. (After all, one could argue that NSF funding is implicitly understood to fund a professor for thinking about other things, as well as for general education of graduate students. [Note that NSF would never admit to the first, though it does to the second.].) My point is only that it's not so outrageous that a political commenter might find the while way the grant system is run to be suspect.

Matt Welsh said...

I should chime in here since I'm one of the PIs on the RoboBees project.

(By the way, I think it's unfortunate when people hide behind posting anonymously. We have nothing to hide here, so why not post using your real name?)

Anyway - the RoboBees project is funded by an NSF Expeditions in Computing grant which is intended to open up new areas for CS. To do that takes a significant infusion of cash. Keep in mind that $10M is nothing compared to the kind of budgets that, say, high-energy physicists get for their fancy new toys to study particles that we're not sure even exist. (I'm being tongue-in-cheek here, in case you can't tell -- the point is that scientific research involves risk, and sometimes significant cost to do it right.)

I don't think we could do RoboBees for the kind of money we typically get for an NSF grant. CAREER barely funds 2 PhD students a year and you're expect to make substantial breakthroughs over a five-year timeframe. It's a joke. Our standards for CS research are really too low since we're used to very constrained NSF budgets. We also seem to expect PhD students to do *all* of the work. To do something like RoboBees you really need full-time research staff and hardware engineers -- people who can build and support things over the lifetime of a project. PhD students can't (and shouldn't) spend too much of their time on the mundane stuff that it takes to make something real. (I mean *really* real. Not "grad student real.")

The bulk of the RoboBees budget is devoted to postdocs and grad students. We have a sizable equipment budget as well, which is actually being matched dollar-for-dollar by Harvard. To do things like build bee-sized flapping wing robots you need more than a couple of PCs and a whiteboard. This is not your typical CS focused project as it includes substantial mechanical and electrical engineering, not to mention neurobiology and entomology components.

The fact that CS folks want to attack us for getting so much funding to do a very cool project is sad. We should be pushing NSF harder to fund more projects like this and make the pool bigger so we can make real breakthroughs. Otherwise all of the money will go to the physicists who (a) know how to tell a good story about the impact of their research, and (b) support each other, to the death, for getting more money. A rising tide raises all boats, after all.

Matt Welsh said...

At least at Harvard, the overhead charged on grants, which goes into the "university", is not folded into the endowment. It's used as one component to pay the substantial operating expenses of a large research university. You know, little things like the electricity bills, machine rooms, administrative staff, etc. This stuff is essential for anyone to get any research done.

The endowment is really a large collection of fairly fragmented accounts, many of which are fairly restricted in how they can be used (e.g., chaired professorships; supporting a library; undergraduate grants; etc.)

Believe me, large grants are not going to line the coffers of the university. This money is pretty much spent as soon as it comes in.

Michael Mitzenmacher said...

Thanks Matt. I was hoping you'd chime in and set the record straight. (As a PI, it sounds better when you do it.)

Anonymous said...

"(By the way, I think it's unfortunate when people hide behind posting anonymously. We have nothing to hide here, so why not post using your real name?)"
-- I am afraid, that professors might form an opinion based on my comments and this might hurt my prospects, when I graduate.
"Otherwise all of the money will go to the physicists who "
-- This does not justify that, CS people also start bluffing the government. We all know that large grants donot fill in the coffers of the university, but they do play an important role in day to day affairs of the university.

"My share of it is about $125k per year. Of that, $60k goes graduate student stipends + tuition. $30k of it goes to the university and my department to cover overhead expenses -- everything from power to administrative support to building maint, etc. About $10k pays for faculty member time. $7.5k goes to equipment. About $7k per year goes to hire undergrads for the project. $6k per year goes to travel. "
This infact proves my point. Universities are charging huge amounts of tuition, saying that they need this for the operational cost, but in fact 120K out of the 1M grant is taken by the university only for building maintenance. Also, the stipend of grad students at CMU is around 2.2K, so infact the rest of 34K USD per grad student is also kept by the university. You can do the maths now, but it seems to me that the biggest chunk of these grants goes to the private universities.
(And yes, physicists are dishonest (if they do the stuff, what Prof Welsh mentioned), but that does not give the right to the CS profs to also behave like their counterparts in the physics department.

David said...

*really* real, not *grad student* real made me chuckle.

Matt Welsh said...

I am not at ALL suggesting that physicists are dishonest. Your reading that into my comments suggests you have a pretty skewed view of how things work.

What I am saying is that physicists are very good at justifying large equipment and operational expenses for research projects at enormous scale, something CS people are not good at. Take the LHC for example: $9 billion all told, although the bulk of that is not coming from NSF. As a community, I think physics has its act together in selling what they do to Congress and funding agencies to get a bigger pot of money to do the work. In CS we have this idea that experimental research can largely be done with off-the-shelf PCs and cheap grad student labor alone, and I think that sells us short when it comes to going after big problems.

But, if you want to keep complaining that CS research is overfunded, I guess you'll have to wait see what it's like when you're on the other side and competing for grant proposals with a 5-7% funding rate and only get half the budget you ask for. Grad students have it easy as you don't generally have to worry about where your funding comes from, so it's easy to complain from where you're sitting.

Anonymous said...

"Your reading that into my comments suggests you have a pretty skewed view of how things work. "
-- No not at all. At one point you say that physicists are very good in justifying all the money that they need, and at the other point you say that, they do this stuff for particles that donot even exist. If this is not bluffing, than I donot know what it is!!!

And please note that, I am not blaming you for anything. I know life is tough for a professor, as the funding is limited and he has to support a lot of people. I just think that this whole system is flawed and the fact that you are unable to see the flaws in the system is really disturbing.

Matt Welsh said...

Now I'm sure you're just trolling and are not serious. Did you not notice that I qualified my joke about nonexistent particles with "I'm being tongue-in-cheek here", or do you not know what that means?

As for "flaws in the system", you seem to have a vague and incorrect understanding of how scientific funding works. You've had three professors trying to correct this misunderstanding to no apparent effect, so I give up. Good luck to you!

Anonymous said...

"Good luck to you!"
--- Good luck to you too...
(By the way, it is okay to accept that one can be wrong some time)

Unknown said...

Media sources compile lists like this all the time. The fact that it's so long should indicate that either (a) Hannity et al. spent several man-days worth of effort to come up with a blog post or (b) it's not quite that thoroughly researched. It's like the "$640 toilet seat" that so outraged the media twenty-some years ago; as I recall, it was actually a custom- built battle-hardened device for an airplane, a device constituting most of the on-board toilet and plumbing. (In retrospect, $640 was a steal.)

The larger point is that in the U.S., as in many countries throughout the world, we spend a lot more money than we have, and a lot of it on things that might not be terribly necessary. It's easier to point out lots of small things and think we can save money by some special commission or some line-item veto. But it's the big-ticket items that get us.* The problem is that some people are more impressed with a 100-item list of million-dollar babies than, say, $780 BILLION spent on the stimulus or a similar amount "invested" in the bailouts. In some sense, that means the tea party movement is a hopeful one; those people actually understand numbers and no longer buy that fiscal responsibility comes from cutting by nickel and dime.

* That's not to say small items don't matter. If a member of Congress is willing to vote for a trillion-dollar spending initiative in exchange for a million dollars worth of spending in his or her district, then you'd better believe we should get after that. But Hannity's list isn't about earmarks, just about spending in general.

Anonymous said...

Hasn't Professor Ron Fearing's group been doing a similar project for years?