Bill asked me if I thought the ICS conference truly was innovative, and in particular how I thought the content compared to that of STOC or FOCS. I've never been to either STOC or FOCS (though I've read some papers and seen some videotaped presentations from those conferences), so I don't consider myself qualified to answer that question directly. However, something related has been on my mind, and I think it's important enough to share with the larger community.
I do believe it is innovative and politically significant that ICS literally represents another country heard from -- and that the derivatives paper appeared there, not in either STOC or FOCS. Compare the derivatives paper to Gentry's homomorphic encryption paper. Gentry's result is of course a stunning breakthrough in an area that had remained wide-open for many years; and, to my (brief) reading, it contains more profound mathematics than the derivatives paper. However, it's quite possible that the derivatives paper will spark changes in the regulation of the multi-trillion-dollar financial product industry. If that happens, it would be reasonable to argue that the derivatives paper would be one of the most influential TCS papers ever.
That comparison, to me, captures the value new concepts can add to the field. US consumers would only have to save $10 million on financial services for Uncle Sam to be 100% repaid for his investment in an Intractability Center. I don't it's a coincidence that that Center's director is a co-author of the derivatives paper, and also a co-author of this CACM position paper on how computer scientists should represent their field to better raise money.
I've had the last two paragraphs of that paper on my office door for a few months now, because I got sick of people complaining to me that there was nothing to be done about financial woes. I'll reproduce those paragraphs here.
One wonders if the failure of computer scientists to articulate the intellectual excitement of their field is not one of the causes of their current funding crisis in the U.S. Too often, policymakers, and hence funding agencies, treat computer science as a provider of services and infrastructure rather than as an exciting discipline worth studying on its own. Promises of future innovation and related scientific advances will be more credible to them if they actually understand that past and current breakthroughs arose from an underlying science rather than from a one-time investment in 'infrastructure.
It is high time the computer science community began to reveal to the public its best kept secret: its work is exciting science -- and indispensable to society.
I also believe it is no coincidence that both co-authors of that paper are on the Steering Committee of ICS. There's nothing like a conference that encourages innovation to demonstrate "promises of future innovation and scientific advances."
A generation or two ago, aerospace contractors used the Space Race as a fundraising tactic. I got the impression from some comments on my first ICS blog post that people were threatened by the idea that China might be a major TCS player, and would prefer if I hadn't even mentioned the possibility. I think that attitude is foolish, and, rather, China's presence on the world TCS stage should be embraced, and used as a reason it is that much more important for the US to invest in theory. After all, if Washington allows things to continue as they are, in ten years, it could be facing a Square Root of Log N Gap!
Okay, that last phrase made me laugh when it popped into my head, so I figured I'd share it. My point, however, is a serious one. A handful of Ph.D's will get postdocs this year at IAS or through the Simons Fellowship. Most people won't, even if they're good. If the CI Fellows program isn't renewed, that means pretty much everyone else is going abroad. In fact, when I was at ICS, a senior researcher told me that he was advising students and recent grads to go abroad, not just for postdocs but also for assistant professorships, and only to return to the US for tenure. That is not a recipe for maintaining scientific prominence in a field, especially if one's "major competitor" is investing heavily in recruitment of theorists.
I will end with a question to consider, if I may. How can we better communicate that computer science, and, in particular, theoretical computer science, is indispensable to society? The government of China doesn't seem to have any trouble understanding this. What about the government of the United States?
There is a HUGE number of postdocs in TOC this year--much much more than I've ever seen. So I don't think people will be going abroad this year. Maybe later when they want an Asst. Prof job.
ReplyDeleteThe CI fellows program did not create jobs--it gave postdocs (at a much higher salary) to people who already had jobs, so that is not going to solve any problems.
Also, even in previous years when not only assistance profs but also postdocs were hard to come by, by and large people did not go abroad. People instead went to the Google and Yahoo and there is nothing that shows this trent changing ..
I have serious doubts that China is supporting ITCS because they find it "indispensable to society." Having grown up in a communist society, I have good reason to suspect that they are doing it for the sake of competition and national pride. It doesn't really matter what the institute is studying, as long as it's doing something that the West is also doing, and they can engage in competition. The success of ITCS is the same as the wild success of communist countries in sports -- certainly the leaders of these countries are not supporting Olympiad-grade gymnasts because they will be great workers in factories.
ReplyDeleteFor us, it doesn't matter why people choose to put money in TCS, as long as they do. But let's be realistic about which selling points work in which places.
If the CI Fellows program isn't renewed, that means pretty much everyone else is going abroad.
ReplyDeleteOr into industry. It is also to our detriment as a field to consider industrial research a second-class option, and to consider those that go there in any way failures.
It is also to our detriment as a field to consider industrial research a second-class option, and to consider those that go there in any way failures.
ReplyDeleteThere are two separate ideas that are often conflated: 1) going to industry is a failure because industry is inherently "lower-class" than academia (an straw man argument if I ever saw one), and 2) going to industry is a failure because one chose to devote 6 years to a Ph.D. solely as a means to do research as a career (note that TCS research is done almost entirely by universities), and knowing in advance that this is unlikely or impossible, would never have entered grad school in the first place and would have gone to industry immediately after undergraduate.
The question is not, "Can a recent Ph.D. go to industry?" or "Should that be considered honorable?" The correct question is, "Do most of these industry jobs taken by Ph.D. graduates really require a Ph.D.?" With jobless Ph.D. flooding the market, of course companies can afford to say, "We will only hire Ph.D.'s for this software engineering position." But assuming that the utterly broken graduate school system in the US were to stop flooding the system with 20 times as many Ph.D. students as there are faculty positions, would these companies go under for lack of sufficiently trained talent? Would they really not be able to find someone who can do the job, because you really need Ph.D.-level training in theoretical research to do so many of these industry jobs? I honestly don't know but I suspect not. Maybe a lot of training is needed for some of these jobs but not in the form of publishing a dozen papers in theory conferences over the course of half a decade.
In that case a Ph.D. taking such a job is a failure not compared to the Ph.D.'s who got academic jobs, but compared to the B.S.'s who have 6 years of additional industry experience/salary raises/retirement money/etc because they didn't waste time in grad school training for a different job than the one they took.
Your post is self-contradictory and a tad bit too much of a rant for my taste. On the on hand you quote a paragraph saying that we should communicate past breakthroughs to a wider audience. On the other hand, you laugh about \sqrt{\log n} gaps. One of the community's past breakthrough results (by a co-author that you mention over an over again in your article) was precisely a factor \sqrt{\log n} improvement. Understanding the remaining "\sqrt{\log n}" gap remains a central challenge.
ReplyDeleteAs pointed out by many others, post-doc offers are looking good this year. See Gasarch's previous post.
"That is not a recipe for maintaining scientific prominence in a field, especially if one's "major competitor" is investing heavily in recruitment of theorists."
ReplyDeleteChina is still spending much less than the US on CS theory. (I don't have numbers, but isn't this clear? Compare the output of China-based theorists to those in the US. I would guess at least ten times less, even though US GDP is only 3.3x China's GDP, and China's population is 4.4x ours.)
Also I think that the current US model of training many PhD students and hiring many postdocs, but then hiring few if any tenure-track faculty, works very well. We get great people spending the best years of their lives working for pennies. When they tire of this deal, they generally go to work for US companies, and do not move abroad as you assert.
why do people say that there are too many PhD's? It's strange b/c it is obvious that the US is behind the rest of the world in math/science/technology and yet people say there are *too many* phDs! Maybe there are not enough jobs! Why should the decision to get a PhD be decided solely by academic demand, when some of those academic demands are not even reasonable, e.g. "lower" rank schools hiring people with Education PhD's over science PhD's to teach science courses.
ReplyDeleteI do think that even if a person ends up in an industry job that they could have gotten with only a BS, it is likely they are more flexible and the BS only person is more specialized for the particular job in which they spent 6 years getting experience.
Anyway, I think the US needs more science research, not less.
"it is obvious that the US is behind the rest of the world in math/science/technology"
ReplyDeleteThis is incorrect. Grade-schoolers may score poorly in the US, but US math and science is at the top of the world.
Yes, I was refering to US citizens who are educated in the US. They are not at the top of the world wrt to technology. That comes mostly from people who are educated for the first 20 years of their lives outside the US.
ReplyDeleteAgree with Mihai.
ReplyDeleteQ: How can we better communicate that computer science, and, in particular, theoretical computer science, is indispensable to society? The government of China doesn't seem to have any trouble understanding this.
ReplyDeleteHmmm ... is the reason self-evident? :)
I think the leader of China has a degree in engineering, so he may understand the importance of math and science education. Also, it's a different culture in the East, where scholarship is revered and introversion is a virtue.
ReplyDeleteAnonymous 12:30: Have you met any industry researchers? I know many, and I assure you they are doing Ph.D.-level research, not just software engineering. I'm not only talking about what Google does.
ReplyDeleteAs pointed out by many others, post-doc offers are looking good this year.
ReplyDeleteThat's a fair criticism. I wrote this guest post before the "postdoc post" appeared, and I didn't have all that data, or I would have modified what I wrote. My real concern, however, is about long-term trends. Just a week ago, the Faculty Senate at my university voted to change the Faculty Handbook in order to create a "rational process" to terminate tenured faculty. This was a response to a university budget crisis and dire state budget predictions.
The national academic crunch is hardly something that an increase in the number of postdocs can fix. Such a "strategy" only postpones the inevitable, by creating glorified temp jobs. Now that a co-director of this blog is on the CCC, it seems reasonable to discuss how best to attract money to the field, not just to comment that things aren't *quite* as bad as I first said.
... A generation or two ago, aerospace contractors used the Space Race as a fundraising tactic ...
ReplyDeleteThe above statement oversimplifies a very complex process, in which theoretical computer science (TCS) played a leading role!
A book (of few) that describes this process in-depth is Neil Sheehan's recent A Fiery Peace in a Cold War, in which we read:
"While von Neumann still kept his hand in at pure mathematics by doing an occasional proof, he had long since become bored with the abstract realm of mathematical research.
He was instead dedicating his nonpareil mind to the practical applications of mathematics and mathematical physics to the service of the American State, first during the Second World War and now in its contest with the Soviet enemy.
With the exception of the Coast Guard, no American military or intelligence organization existed that John von Neumann did not advise."
The point is, the vast enterprises of the 1950s did not exploit TCS ... they were catalyzed-by and organized-by TCS.
So one might reasonably ask, what similarly vast enterprises can be catalyzed by *today's* TCS?
You seem really overly impressed with this derivatives paper.
ReplyDeleteIn that case a Ph.D. taking such a job is a failure not compared to the Ph.D.'s who got academic jobs, but compared to the B.S.'s who have 6 years of additional industry experience/salary raises/retirement money/etc because they didn't waste time in grad school training for a different job than the one they took.
ReplyDelete1. I don't believe it is true that if you join industry after getting a PhD, you do the same work that someone who got a BS and worked for five/six years does.
2. Even assuming for the sake of argument that they do the same work, you do get a higher pay. I don't see how enjoying five/six of grad school education, learning stuff you are interested in, doing things you love doing and then joining industry is "failing" compared to getting a BS and joining industry. Maybe your metrics of success are different.
Have you met any industry researchers? I know many, and I assure you they are doing Ph.D.-level research, not just software engineering. I'm not only talking about what Google does.
ReplyDeleteAnd I have met others who do not do PhD-level research. The relevant statistic is not, do there exist some of each, but how many truly PhD-level jobs are there? How many PhDs are there? What is the ratio of the first quantity to the second? If it is significantly less than 1, then that is a problem.
Even assuming for the sake of argument that they do the same work, you do get a higher pay. I don't believe it is true that if you join industry after getting a PhD, you do the same work that someone who got a BS and worked for five/six years does.
It's not "true" or "false" in general; it depends on the job. Some jobs require PhD-level training so if you take them, you will do different work for higher pay than a BS. Again, the correct question is, how many such jobs are there, and how many PhD's are there? Unless most of these jobs forbid publication for intellectual property reasons, we can safely state that the number of industry jobs involving theoretical research is far less than the number of academic jobs doing this research. Why can we conclude this? Go through STOC/FOCS/SODA proceedings and count how many authors are from industry. It's a tiny minority. It seems reasonable to state that focusing primarily on publishing in these conferences for 6 years is not appropriate training for a job unless that job involves doing the kind of research that gets published in these conferences. Is this reasoning unsound? Conversely, if you consider there to be vast swaths of "legitimately TCS" industry jobs, but those people are not publishing in the main theory conferences, then the conclusion is that those conferences are not legitimate TCS.
I don't see how enjoying five/six of grad school education, learning stuff you are interested in, doing things you love doing and then joining industry is "failing" compared to getting a BS and joining industry. Maybe your metrics of success are different.
ReplyDeleteReally? You don't see how some people (though not all) could find a downside to this? It's one thing to say, "I personally don't mind this career path because of all the positives." But to state that there are no negatives, or that everyone prioritizes things the same way as you so that they would also not mind the negatives, is downright blind. There are many disadvantages to grad school: You make less money. It is much harder to start a family (both due to money and due to time constraints compared to a 9-5 job). You can put away little or no retirement savings until you graduate. For a lot of those industry jobs (picture a small startup) it is a real advantage to be 22 years old, unmarried, with no kids and lots of energy, so in a narrow sense (but perhaps important to some companies) you are *less* qualified after grad school than before. I'm honestly happy for people who are above the fray and not bothered by this stuff. But I am, and I don't think that there is anything wrong with that.
Clearly, those of us who go to grad school do it because we decide that these negatives are outweighed by the positives, but the positives are different for everyone. Even though some (like you) are happy to get a PhD just for the education and self-fulfillment, I have to believe that others (like me) do it primarily because we have a particular career in mind (research) that we are not allowed to do without a PhD, so we accept the negatives solely as a means to obtain this goal. For me, to get a PhD and not do research for a living is like going to medical school and not being a doctor. This is a perfectly legitimate opinion even if you don't share it.
Both of the last anonymouses make it sound like someone FORCED them to go to grad school. If the downsides are so bad, then don't go!!
ReplyDeleteOn the other hand, if someone WANTS to go and someone else wants to advise them, what does that have to do with you?
I don't think you are entitled to a career doing research just because you earned a PhD. When one goes to grad school, one should look at the downsides.
ReplyDeleteA comparison can be drawn to NFL; there are thousands of college football players but only a few hundreds will ever get a chance to play at the pro level.
In early 20th century, WWWII has given America a chance to welcome all the best european minds, including aforementioned von Neumann, or Einstein. Wise government has given this brilliant people possibility to work not just as a dishwashers, but continue research. Today America is No1. But the world is changing (yes, it can), and the story repeats. Now other countries can offer a better deal for researches and open their doors wide. Bad management kills any good business. So my strong believe is no matter how hard US tries (except for war), in the next 20 years we will see a generation of 1st class chinese scientists educated by current MIT etc PhDs in China. And then, it will not be as easy to get a position in China for the same MIT etc graduate anymore. So, I would rather go there now :) Especially, after all the negatively colored discussions seen on this blog. US has to try harder.
ReplyDeleteUnless most of these jobs forbid publication for intellectual property reasons,
ReplyDeleteNearly 100% of them do, plus if you are busy making something work in real life, usually you can't spare a couple of weeks to dot the is and cross the ts to get the results ready for publication.
It seems reasonable to state that focusing primarily on publishing in these conferences for 6 years is not appropriate training for a job unless that job involves doing the kind of research that gets published in these conferences.
You seem to have very strong opinions for a person who's never been in industry.
The relevant statistic is not, do there exist some of each, but how many truly PhD-level jobs are there?
Even this is an incorrect statistic. From what I hear many PhDs working at a certain large search engine company are not doing PhD level work. This company is so large and such an attractive employer that it will skew the statistics. But the point is that PhDs choose to work there. They the could have gone to any number of other companies instead in which their hard won PhD talents would have been put to good use, not publishing papers in STOC/FOCS/SODA, but gasp, actually doing relevant, world changing work.
As chair of a search committee in CS from a fairly large public university this year, I found it an interesting comment on the US education system that less than 15%of the nearly 200 applicants for our tenure track position were US citizens.
ReplyDeleteYou seem really overly impressed with this derivatives paper.
ReplyDeleteI agree. Also, this comment reminds me of a disparaging remark I heard at a talk by a rather famous combinatorialist about logic, which equally well applies to theoretical computer science. Adapted to TCS the remark would become: "Theoretical computer science is about as useful in real life applications such as derivative trading or cryptography etc., as the "Wrong way" signs are to highway driving."
Can someone please explain why they think the derivatives paper is not applicable in real life?
ReplyDeleteTheoretical computer science is about as useful in real life applications such as derivative trading or cryptography etc.,
ReplyDeleteLast I checked Theoretical Computer Science includes things such as B-trees, hashing, streaming algorithms, optimization algorithms, auctions all of which are useful in real life.
Noam Nissan's SODA talk was a nice example of this. The algorithms developed by his group are running google TV ad auctions.
There's a tendency here to identify your own subfield of TCS with all of TCS, but this is just not so. There are at least three very distinct flavor of communities within TCS (in alphabetical order): Algorithms, Complexity and Theory B. Be careful with generalizations.
There is a HUGE number of postdocs in TOC this year--much much more than I've ever seen.
ReplyDeleteHmmmm ... taking the viewpoint of a Dean of Engineering who is looking at a dismal budget sheet ... and yet who wishes to sustain their school's capability for engineering innovation ... is this because, of all engineering academic positions, a TOC postdoc (arguably) requires the *least* resources?
After all, expensive new physical infrastructures like wet-bench labs, nanofab facilities, wind tunnels, supercomputing facilities, materials science labs, etc., are not realistic options for many engineering deans nowadays.
Short-term, this is (perhaps) very *good* news for TOC ... long-term, it is (perhaps) sobering news in relation to the health and vigor of the overall science-and-engineering ecosystem.
As Aaron says, postdocs are glorified temp jobs. There are a huge number of postdoc positions and next to no tenure-track positions. The field is training far too many PhDs. This is hurting a lot of people. I do not know how much it is driving quality down by pushing good people out and leaving the sloggers. However, the environment is poisonous these days and I think this is hurting the field.
ReplyDeleteHowever, it's quite possible that the derivatives paper will spark changes in the regulation of the multi-trillion-dollar financial product industry. If that happens, it would be reasonable to argue that the derivatives paper would be one of the most influential TCS papers ever.
ReplyDeleteYeah, and when we meet aliens with PSPACE-capable brains, we'll be awfully glad we know that IP=PSPACE, so that the aliens can share their insights with us. Then Shamir's paper will be seen as the most influential TCS paper ever.
The derivatives paper is a nice idea, but nobody's ever going to look back on it as the paper that led to a reworking of financial industry regulation. It's exaggerations like this that make "conceptual" research come across as silly. It's not enough to have a good idea - instead, everything has to be an astonishing breakthrough that will soon revolutionize everything.
Having spoken to a relatively large number of Chinese young people, I am under the impression that they tend to find theoretical CS useless in the same proportion as their western counterparts. And not only that, they also have a lack of appreciation of doing things for the sake of themselves. I rather think that the reasons the Chinese government is supporting ITCS are the ones pointed out by Mihai (and being more precise, because the only Turing Award from China is in that area).
ReplyDeleteRegarding the percentage of foreign students in the US, I think it says more about the US ability to attract the best students in the world (and how the world is getting better at CS and science&technology in general) than anything else.
I also agree with the commenter that pointed out that people seem to assume that everyone else has the same priorities in life. Six years spent learning stuff that you find interesting, surrounded by people that you think are interesting, are not something that I would consider wasted. I can see why someone with different values that mine would do otherwise, but I would appreciate it they did not extrapolate that to other people, and say that they are wasting their years.
"Regarding the percentage of foreign students in the US, I think it says more about the US ability to attract the best students in the world (and how the world is getting better at CS and science&technology in general) than anything else."
ReplyDeleteReally? There are whole top-ranked departments in which there are maybe one or two Americans. That doesn't say anything about the quality of science education Americans are getting?
As a Chinese researcher working in a NON-Tsinghua university in China, I want to say that Yao's group is so well supported because they are doing good stuff. And for what reason do you think it is just TCS gets particularly well supported? Take a look at other fields, there are many other research groups, such as Professor Yigong Shi in Biology, who was recently appointed by Tsinghua and also got well supported. I think the main reason that the government supports these people is that THEY DO GOOD RESEARCH. I don't expect the funding agencies love my area like I do or believe they are useful even if they are not. But I fully support that they gives good supports to the people who are doing good research.
ReplyDeletePrejudice? Hatred? Judging before knowing what is really going on? You guys are very disappointing and embarrassing the name of scientists.
I want to comment the previous poster (from China), and to comment further upon their topic of providing "good support to people who are doing good research."
ReplyDeleteIn the prosperous period from (say) 1970 to the present, perhaps 4x10^8 people lived in societies whose young people could reasonably choose to pursue a career in research.
In recent years, that number has grown immensely; presently it may be as large as 2x10^9 ... which is a quintupling in a relatively short time. This is amazing ... and it is *good* news IMHO. And by the end of the 21st century, this number is destined to quintuple again ... we may hope!
But is it any wonder that math, science, and engineering are feeling some strain?
To accomodate this increase will require some imaginative thinking. Right now, about 1/10^4 American citizens is an AMS member.
If this incidence stays constant during the coming expansion, the planetary population of AMS-equivalent mathematicians will be about 10^6 (wow! ... I want to be in Las Vegas when this meeting is held).
How will these 10^6 mathematicians keep busy? How will they respect traditional values and enterprises? What new values and enterprises will they create?
These are IMHO the most interesting and enjoyable---and challenging---questions that Aaron's post suggests.
To the Chinese poster: People are just arguing about facts, not to be offensive. Take it easy and sorry if it hurts you.
ReplyDeleteBTW, I am Chinese too.
People are just arguing about facts, not to be offensive. Take it easy and sorry if it hurts you.
ReplyDeleteThe only fact here is that the Chinese government is investing substantial amounts of money into TCS. Everything else, about possible motivation etc. are pure speculation (no one here as any direct contact with Chinese govt officials who are making these decisions) -- some coming from disgruntled ex-East Europeans who believe nothing good can come out of a communist govt. (In fact, these people should read the article appearing in the New York Times today about how because of government investment China is racing ahead in the area of renewable energy -- and how the West will soon be reliant on Chinese technology in this area.
I fully sympathize with the non-Tsinghua researcher here -- instead of lauding the Chinese government for making wise investments for the future, and trying to see what we can do to convince our own government to do the same, more people seem to busy trying to find ulterior motives etc.
The Chinese government, run by engineers, instinctively knows that you can't go wrong investing in scientific knowledge even if it's for its own sake. Somehow and someway the investment will pay off handsomely in the future. It will only be good for the country.
ReplyDeleteThe U.S., however, takes a much different approach to such investments. When the going gets tough, guess which subjects are the first to be "de-funded"? That's right, the cheapest ones to support -- the theoretical ones. For example, some years ago the Ph.D. program in mathematics at the University of Rochester was recommended by university officials to be eliminated because of budgetary problems. It took many persuasive letters from academic mathematicians and scientists including Nobel laureates to convince them to reverse their decision.
Unfortunately, I think we might have to suffer the long-term consequences before U.S. decision makers will finally realize the disaster of not investing in knowledge.
Can somebody other than Aaron Sterling who has more experience in the field write a guest post on ICS ? There were so many big shots present there, Sanjeev Arora, Sanjeev Khanna, Cynthia Dwork, Avrim Blum, Bernard Chazelle, Shafi Goldwasser and many more (check http://conference.itcs.tsinghua.edu.cn/ICS2010/).
ReplyDeleteGasarch, can you please request one of them to compare the content of ICS papers with STOC/FOCS/SODA ?
Why do you need big shots to tell you what to think?
ReplyDeleteAaron himself said he has never been to STOC/FOCS and thus he thinks he is not qualified to answer Bill's question. The big shots I mentioned were present at ICS, supposedly heard the talks and have published at "many" FOCS/STOC/SODA. So their comparison will give us more idea what to expect from ICS in future.
ReplyDeleteWith regard to "Should one ask the big-shot experts?", the verdict of history is that the big-shot experts will very often give correct answers ... to what will very often turn out to be the wrong questions.
ReplyDeleteSo perhaps the question that the post asks, namely: How can we better communicate that computer science, and, in particular, theoretical computer science, is indispensable to society?
... might possible morph into ...
How might computer science, and in particular, theoretical computer science, become more evidently indispensable to society?
The Chinese government, run by engineers,
ReplyDeletehahah ... wat a joke ... u make me laugh aloud when reading ur insights about the chinese government.
Hello. I believe that going abroad is as useful as essay writing because both make you think about the relevance of this action, analyze and summarize info, and conclude. I have found the article is interesting as well as most comments.
ReplyDelete