My painful lesson for the week
Years ago, Sasha Razborov taught me one of my all-time favorite jokes.
In the 1960s, a man starts handing out leaflets in Moscow’s Red Square. Needless to say, he’s immediately apprehended by the KGB. On examining the leaflets, however, the KGB agents discover that they’re just blank pieces of paper. “What is the meaning of this?” the agents demand.
“What could I write?” exclaims the man. “It’s so obvious!”
The lesson I’ve learned this week is that the man was wrong. In politics, nothing is ever too obvious.
Follow
Comment #1 December 18th, 2010 at 11:18 pm
This is a blank comment too:
” “
Comment #2 December 19th, 2010 at 12:47 am
What is this a reference to?
Comment #3 December 19th, 2010 at 1:58 am
The comments on the QCut post were enlightening for me. At first I totally agreed with the post (the sarcasm was obvious). But then one of the commenters pointed out that you yourself got $500,000 (!) from NSF. Are you planning to actually build and maybe mass-produce some real physical quantum computers? If not, I can hardly imagine what one researcher in quantum complexity needs 500k for, even if it is spread out over 5 years. Are you hiring five grad students who would otherwise not be able to get stipends?
Comment #4 December 19th, 2010 at 3:20 am
IMHO, you misquote Sasha:
The date is certainly later than 1950s (KGB was formed in 1954), and the “event” ought to be happening on the Red Square (that’s the most usual setting for this joke, certainly).
(FYI, Moscow has lots of train stations, including half a dozen huge train terminals, so “the train station” simply makes no sense…)
Comment #5 December 19th, 2010 at 5:26 am
NSF grants cover a portion (2 months, or 1/6) of the PI’s salary. So that accounts for a large fraction, maybe $120K. Then you have grad students who in CS can easily cost $60K a year each (stipend plus tuition plus benefits). I don’t know MIT’s tuition. Then you have equipment and travel for everyone supported. If there is anything left over, it can be used to support a postdoc for a year or two.
It does add up.
Comment #6 December 19th, 2010 at 6:01 am
Dima: Thanks—I’ve corrected the joke!
Comment #7 December 19th, 2010 at 6:20 am
Anon grad student: I see that Bill already partly answered your question, but just in case you’re curious…
Of the $580,000 award, more than half is immediately taken by MIT as overhead tax, where it goes toward funding MIT operations in general. Of what remains, most goes toward funding
(a) my three PhD students (yes, they apply for and sometimes get their own stipends, but those are harder and harder to come by),
(b) a postdoc who’ll be starting with us next fall, and
(c) my summer salary (which, following the standard practice, MIT doesn’t pay).
In addition to the CAREER, I’m lucky to also have DARPA and Sloan grants and some remaining startup funds, and also to be part of an NSF STC (“Science and Technology Center”) on the science of information. The CAREER is the majority of it, but it wouldn’t cover everything.
Comment #8 December 19th, 2010 at 8:34 am
To explain overhead: the 50% or so that MIT takes off the top goes to pay for costs like office space, utilities, and services. Neither Scott nor his PhD students pay directly rent on their office space, utilities to keep the lights on, salaries of departmental secretaries, administrators, building facilities managers, or custodial service to keep the building operational. A big chunk of the money goes to the library whose journal subscriptions fund scientific journals that make the results of Scott’s research available to the general public (including any future commercial quantum computing manufacturers).
Some of this is paid for by tuition, the endowment, and income MIT generates from their share of patent royalties, but a lot of it is paid for by grant overhead. The situation is even more serious at public universities which have lower tuition and are facing huge cuts in their state subsidies due to bankrupt states governments and the science departments have to compete for university general funds with departments that generate less revenue internally.
50% seems like a lot, and it is — professors love to grumble about taxes as much as the next citizen even when they understand the reasons for it. Overall, Scott and the NSF get a fairly good deal for the overhead paid, compared to what all the services would cost him — not to mention the intangible value of actually working at MIT where if he reaches a stumbling block, he can just walk down the hall and probably find a world expert on the subject. Overhead for corporate R&D or at national labs is usually considerably higher than academic institutions.
Comment #9 December 19th, 2010 at 11:10 am
What is this a reference to?
Thank you for illustrating the lesson for me! 🙂 It’s in reference to the QCut post.
Comment #10 December 19th, 2010 at 11:35 am
Scott, what’s painful about this episode (IMHO) is not that it represents a nadir of the STEM polity’s internal dialog in relation to complexity theory and QIT, but that it approximates more nearly the present-day apex of dialog quality.
Yikes. Complexity theory and QIT among humanity’s best prospects, for creating the STEM resources we need to get through this century … can’t we do better?
Earlier generations of complexity theorists set us an example whose value has endured.
Sarcasm and mockery are uncommon in the CT/QIT Founders’ writings … yet by no means totally absent, Szilard’s The Voice of the Dolphin, is a classic example … perhaps the lesson is that small amounts go a long way.
What is strikingly absent in the present dialog, is evidence that we have learned all that much since the 1950s. As Goethe put it:
Comment #11 December 19th, 2010 at 11:36 am
“A big chunk of the money goes to the library whose journal subscriptions fund scientific journals that make the results of Scott’s research available to the general public”
I’m a fan of massive increases funding of basic science research and research universities, but boy does this part of the “MIT overhead tax” sound dated.
Comment #12 December 19th, 2010 at 12:15 pm
Hopefully Anonymous: Amen! Indeed, previously on this blog, I’ve advocated boycotts of most commercial journals (something that I’ve practiced for seven years or so). The cost of most journal subscriptions has been increasing exponentially, even as their actual value has declined to almost zero.
Comment #13 December 19th, 2010 at 1:23 pm
It’s totally disingenuous to include a grad studen’t tuition waiver as part of the expense: they are probably post-candidacy and not taking any classes besides “independent studies” with… you! What is their take-home pay? Probably about $20,000/year or less. Do benefits really cost $40,000/year? Also, totally agreeing with Hopefully Anonymous: commercials journals are nothing but a rent-seeking middleman, and ought to be replaced by online systems with a fraction of the cost. Who uses paper nowadays anyway? Certainly, when I myself publish, I’d much rather my papers be available free to the public, rather than behind an outdated paywall. –It kind of goes against the very meaning of the word “publish”!
But besides those two points, thanks for the clarification.
Comment #14 December 19th, 2010 at 1:39 pm
Incidentally, I know some out-of-work physicists with bachelors degrees who, for $500,000, could probably make real strides toward an actual QC, rather than publishing obscure papers behind paywalls about conjectures about whether certain subclasses of quantum computability are equal. Even if they failed, the endeavor would have a much more stimulating impact on the economy than giving half of it to MIT.
Comment #15 December 19th, 2010 at 1:57 pm
Anon grad student: I think you’re uninformed or misinformed on several points.
1. There are lots of in-work physicists with PhDs who get a lot more than $500,000 to work toward building actual QCs! They’ve made major advances, but they haven’t built a scalable QC yet. That’s because the problem is hard. You should suggest to your out-of-work friends that they apply to one of the many labs working on this, if they’re interested!
2. The amount of money that goes toward quantum computing theory is a minuscule fraction of the amount that goes toward quantum computing implementation—probably a few percent (depending on how you measure; the boundary between theory and implementation is a blurry one). Yet, without theorists like Bernstein-Vazirani and Shor discovering the possibility of a quantum speedup in the first place, there would’ve been no reason for any of the implementation efforts.
3. Essentially every quantum computing paper ever written is available for free on quant-ph, if you’re interested to look. So the only question is whether those papers should also be submitted to pricey journals. As I wrote above, I’ve been saying since I was a grad student, to anyone who cares to listen, that the answer is a resounding no! That’s why I support open-access journals like Theory of Computing.
Comment #16 December 19th, 2010 at 1:59 pm
First of all… Those conjectures about certain subclasses is all complexity theory is. Without them, people would still wonder why they can’t find a good algorithm for that problem and they couldn’t even think of building a quantum computer, because they wouldn’t know if it is worth it. Plus, those conjectures are always reasonable and this comes from a person who thinks that there’s a good chance that P=NP but prefers living in the real world.
Plus, you cannot distinguish between sciences now, can you. As it was said in the original posts, every research can be expressed so that it sounds trivial . Classical example of our times is people that don’t even know what a particle is saying that LHC is a big tunnel so that scientists can play ping pong spending billions.
As a final note , I am a student in Greece at the moment. Students and even professors have to pay themselves for the conferences . My department has 800 students and its funding for this year was 1200 dollars (facilities+salaries excluded, but it would be a good idea to take some photos from our “facilities). Yeah, go figure. The whole budget (everything included) got slashed 40% down and professors lost around 8000 dollars. So you should take a good look in this number everyone and instead of fighting each other, work together to come to an agreement about funding, or risk becoming more like us.
Comment #17 December 19th, 2010 at 3:11 pm
I don’t think MIT takes 50%. You put together your budget, then add 50% of it to that to cover overhead. That means MIT is taking 33% of the total.
Overhead rates are formally negotiated between universities and federal research bodies, by the way. It’s not like NSF is just acquiescing to MIT’s demand for money.
For a grant with two PIs, two grad students and two months of faculty summer support plus overhead gets you very close to the upper limit of NSF “small” grants, and you still need to find room to get those grad students to conferences and stuff. They really aren’t that much money in the grand scheme of things. A small business (which is essentially what a lab is) would run through a lot more cash in a much shorter period of time.
Comment #18 December 19th, 2010 at 3:42 pm
Also, think about how much Google is worth to today’s GDP. Just in the sheer amount of time it saves it must be worth billions. But a company like Google would not exist without fundamental research in CS theory, systems, and AI (not to mention other fields like information science, library science, etc) funded by the NSF. At the time it was funded, much of that research would’ve looked ridiculous on its face. And yet it ended up with a huge return on investment to the federal government.
Comment #19 December 19th, 2010 at 6:51 pm
Regarding overhead: It’s not quite as bad as Scott says, although it depends if you count grad student “tuition” as overhead.
Most universities take somewhere between 50-65% “overhead” — my own, Princeton, takes 62%. Which means for ever dollar I spend, I need to pay the university $1.62.
What that means is that a single grad student, once I include their “half” tuition that the university charges (~18K), salary (20K), summer salary (8K), computing fees, and overhead, comes to 75-80K / year. (Sidenote: if you consider tuition as overhead — which is sure seems like it after students stop taking classes — then overhead actually looks more like 150%, which is pretty ridiculous.)
So a CAREER grant for 5 years at $500K (which is standard for CAREER) funds a SINGLE grad student per year, plus $20-25K left over for (1) conference travel, (2) misc (including student laptop, etc.), and (3) 1-1.5 month of PI salary.
If you’ve never tried making “payroll” yourself (which turns out also to be the job of professors), the money goes a lot faster than you think!
Comment #20 December 19th, 2010 at 7:15 pm
lylebot and Mike: Thanks! I stand corrected about the actual overhead rate being r/(1+r), where r is the quoted rate. (For some reason, I’ve always had trouble keeping track of dollar amounts to within more than an order of magnitude. 🙂 )
Comment #21 December 19th, 2010 at 7:27 pm
Very interesting. As a grad student, I’d sure like to have even one half of that 75-80k/year which apparently is being spent on me. In fact it points out some pretty ridiculous inefficiency in the system– which seems to vindicate Eric Cantor (hurts me to say that, being a radical leftwinger). Why can’t Scott and his students write their papers from their own homes? (For those who’d say “because you can’t get grants without affiliation”, how does that make NSF and the universities anything more than a protection racket?)
Comment #22 December 19th, 2010 at 7:44 pm
Anon grad student: I agree with you that there are large inefficiencies in the way grant money is spent, owing to lack of flexibility. However, the obvious solution to the problem that you describe is probably the opposite of what Eric Cantor and his friends would want! Namely: give people who are doing good research a lump sum, then sit back and give them wide discretion as to how much student support to pay, whether to rent office space, etc. (For example, I’d certainly pay my grad students more were the option available.)
Comment #23 December 20th, 2010 at 12:58 am
Argh, I don’t understand the joke! Can someone please explain it to me?
Comment #24 December 23rd, 2010 at 10:16 pm
Clearly someone should get a career grant (etc) in management to study how to make the grant process and associated overhead more efficient. 🙂 I too do wonder about the traditional way of doing things in academia. For example, are conferences really necessary at all in this day of the Web? Still, I think it would be very dangerous to start cutting a system as productive as American advanced education to the bone without a thorough understanding of what one is doing. It’s easy to go “cut this! slash that!” But there really is nothing like what the US does anywhere in the world. It should be considered a national treasure.