Does the Military Dominate CS Research?

An anonymous reader asks: "It seems at my university the military has their fingers in much of the computer science research happening on campus: sensors, intelligent agents, autonomous vehicles, supercomputing. Is this the case at other schools around the US? How about outside of the US? How is the military shaping the current state of CS research? What areas of research atrophy because the funding goes to investigating military applications of new technology?"

Comments

It's not uncommon.

[Anonymous Coward]

That was similar to my experience back when I was in school.

And considering the history of computing, it is to be expected.

Computers are tool for automating complicated by not particularly egaging tasks.

From code breaking, to calculating artillary tables, to distributing information. It's not know how that's the obstical, but maintaing focus and attention. With the millitary, few people have the resources or the motivations to tackle the extremes that remain, besides them. The upside is, while the projects might be defense oriented as far as the money is concerned. The people aren't. Some of the people writing their doctoral thesis based on those projects might just want to make the most kickass games (like one of my CSE TA's). Smarter robots might well lead to smarter monsters.

That's just the nature of the bleeding edge, the inscentive is always going to be strongest for militaries. You can get wrapped up in black helicopters or Chile 1950. Or you can step back and know them for what they are, individual quanta which are part of vast spectrum. While the military might have given us ICBM, and the possibility of nuclear holocaust, the secondary benefits were world wide communications satillites, GPS, the internet, aluminum cans, nuclear power, the death star, and the only chance to defend ourselves from a rogue asteroid.

Be happy for the money. Be happy for the challenge. Be happy for the opportunity to hone your skills.

Which areas atrophy?

[PM4RK5]

My guess would be that the fundamentals of system design atrophy. In many ways the computer science field has seen very little innovation as far as "operating systems" are concerned. At least in the recent past.

Most systems try to mimick windows or something else, except for Apple's OS X. But on the other hand, it is also built atop a UNIX-style system, and is thus somewhat based on old ideas.

The IT industry has created such large barriers to entry that any new or radical ideas as far as desktop systems go (or servers, for that matter) have failed to enter the market successfully. Arguably, Linux's success is due to the fact that it's just a reimplementation of the old UNIX system design.

Colleges and other higher-level academic institutions are the testbed for new ideas in the CS field, and things like system design and a computers' fundamental setup have atrophied over the past few years, since I, for one, have seen very little that qualifies as "new."

One thing I would like to do is try to completely reinvent the desktop system in college as a project, because many, many technologies are just improvements upon older ones. What the industry needs is a radically new system that takes advantage of what's out there now, as far as both ideas go and as far as hardware goes.

This is just my two cents, but if you look at basic system design (device drivers, processing, filesystems, et cetera), there has been very little that is radically new.

This is why I think basic system design has atrophied at the expense of other areas.

Re:Which areas atrophy?

[Phaid]

This is just my two cents, but if you look at basic system design (device drivers, processing, filesystems, et cetera), there has been very little that is radically new.

That's basically because the problem has been solved, and there's just no need to go reinventing that wheel. The approaches to operating systems, whether for desktops, small embedded systems, etc, are well understood, they are tested and reliable, and there isn't a magical new technique which will meaningfully improve on them. The only real "interesting" research in operating systems that still takes place is in very large systems, managing large numbers of CPUs, etc. And given the nature of military applications, I seriously doubt that those areas of research lack funding.

This is not to say that areas where the user interacts with and configures the operating system don't need work -- giving users an easy way to add and remove device drivers, for example -- but those aren't core operating system concepts, those are user interface issues, and it's important not to confuse the two.

Re:Which areas atrophy?

[shaitand]

"there isn't a magical new technique which will meaningfully improve on them"

Nobody can make such a statement about ANYTHING with accuracy. The fact is that if there is a magnificant new technique that improve something, we aren't going to find it if there is no research being done on the subject.

Re:Which areas atrophy?

[Rocky]

Your answer is too simplistic.

Here is an excellent overview of exactly what is wrong with current systems research.

The gist of it is (IMHO): the current research atmosphere is too short term to support a truly revolutionary systems research program - and a good one requires more resources than one can justify. The problems are nowhere near solved - everyone has just settled on some fixpoint.

Re:Which areas atrophy?

[Phaid]

Yeah except that paper really doesn't do anything more than rail against the current state of affairs. So what if 20 years ago people coming into Bell Labs were used to using 20 different operating systems and now they only use one? That doesn't disprove what I said in any way at all. It's basically just a longer and more eloquent way of saying the exact same thing that the original poster I responded to was saying -- quoting this back to me is really just begging the question.

Look, we already know how to make a protected memory, multitasking operating system that runs on commodity hardware and which, for the average user, provides multitasking performance indistinguishable from running individual tasks on a dedicated machine. We already know how to make a hard real-time operating system (and yes, BeOS shows us that essentially, realtime suitability multiplied by desktop suitability is a constant). We already know that microkernels don't do much more than protect you from badly coded drivers, and at a performance cost. None of these things are a mystery any more, and there just isn't a single-processor operating system model that's going to come along and revolutionize OS design on current hardware.

Multiprocessor OS'es? Yes, as I said there's plenty of room for research there. Come up with an analog processor or some other hardware revolution that we don't currently think about? Yes, that would likely turn commodity OS thinking on its ear. But there is simply no interesting innovation left in the nuts and bolts of operating system software for current commodity hardware - all of the interesting research is either at a lower (hardware) level or at a higher (way more hardware) level.

In terms of fostering new research, the one genuinely interesting statement in that paper is:

Only one GUI has ever been seriously tried, and its best ideas date from the 1970s. (In some ways, it's been getting worse; today the screen is overed with confusing little pictures.) Surely there are other possibilities. (Linux's interface isn't even as good as Windows!)

But look! He's talking about user interfaces there, not the core of the operating system. With all due respect to Robert Pike, all he's doing in this paper is expressing frustration that the good old days are over and people aren't doing fun research any more

Face it

[Bluesman]

The reason most things get done is because there is a pressing need, and people in the military understand this better than anyone else. Defending yourself and defending a nation is a pressing need, whether you'd like it to be or not.

It would be nice if we could all hold hands and do research for the good of humanity, but unfortunately, human beings aren't wired that way. Nothing would get done. This is why communist societies, which are beautiful ideas on paper, don't work.

One of the major fallacies that far too many people put a lot of faith in is that people are basically good. We are not.

I know this might sound harsh and cynical, but the fact of the matter is, once you accept the fact that human nature is brutal, selfish, and ugly, you're most of the way there toward a realistic world view.

And if you look at it pragmatically, you'll realize that necessity is the mother of invention, and almost all of the great technological advances in history stem from military necessity.

Yeah, it's not nice. No, I don't like it. But that's how it works.

Re:Face it

[Anonymous Coward]

Serious question: Do you consider yourself to be brutal, selfish, and ugly?

Don't let a few bad apples spoil your view of humanity. The people you described currently hold a lot of sway, but things change constantly. I don't expect a utopia in my lifetime, but I don't expect things to always be ugly either.

Re:Face it

[Bluesman]

That's true, but it's only true because from the time we're born, society civilizes us.

I'm not under the illusion that everyone's a jerk out to get me, I'm just saying that absent a civilizing influence, we're all a bunch of animals. That's why education and parenting are so important.

But it's also why we need to constantly be on guard, because there are always going to be parents who don't do their job, and societies that fail to civilize their next generation.

This is why societies fail. I have no doubt that American society will fail this way. It's happening around you.

We like to laugh at people who point out that our country is going to hell in a handbasket, with the excuse, "Yeah, every generation says that. Parents always remark about how much better it was when they were kids."

But every generation is right. Parents are right. The longer a society lasts, the more apathetic it becomes. It's happening all around you. Americans have a very sheltered and distorted view of the world. We have become a culture of ignorance, apathy, and selfishness.

So while it's understandable to me why someone would look at military sponsorship of research projects and think it's a bad thing, I find the idea ridiculous. This is one of the only ways such things get done, and perhaps one of the best reasons to do it. Advanced technology helps avert war and reduces casualties when war cannot be avoided. It's a fact. Not to mention the beneficial uses of military technology soon follow.

Re:Face it

[Bluesman]

>What makes you trust our generals and technocrats so much?

If you don't trust someone in the military more than you trust an average joe, odds are you haven't spent much time working with the military, if any at all. I have no qualms about saying that people who serve in the military are hands down the best people a country has to offer. That goes for most modern democratic societies. It's why you hear all the anti war crowd still saying "oh, I still support the troops." Why would you support the troops but turn on them when they rise to the rank of general? Is there some transformation that takes place? Generals are not appointed, you know.

The whole idea is not to come up with a better way to kill people, in order to go out and kill them. If that were true, we'd have nuked the rest of the world while we were the only ones with the bomb. The idea is to maintain a military superiority so we don't have to fight, and when we do, we fight as humanely as possible, with as little loss of life on either side as possible.

I know it's difficult for an idealist to accept, but defense, killing, and wars are sometimes necessary. It's best to be prepared. I'm for giving my generals the military might they need to win decisively, because, hell, they're on my side, and I think my side is better than all of the others.

Humanity is not perfect, my country is not perfect, but it's the best there ever was. I think we've accomplished a lot of good and prevented a lot of violence through our military influence. Most recent case in point is Libya.

But hey, I'd be perfectly happy for the US to pack up all around the world, and leave the rest of the world to itself. You can imagine how many days it would take for China to obliterate Taiwan, for Israel to flatten the rest of the mid-east (yes, they have a very advanced military), etc. etc. etc. You can imagine what the Soviet Union would have done absent our influence.

On one hand, it would be so gratifying to sit and watch Europe have to eat its words, on the other, I'm glad our generals are the kind of people they are.

Hell yeah

[Will Sargent]

The military funded the Internet.

The military, as one of the largest software developers on Earth, basically created software engineering and still pushes for hard numbers from projects and code.

When the military tried out OO technologies on flight simulators, they scheduled five different projects, the first one of which was set up to fail(!) so they could accurately determine what actual benefit they could get from OO.

The military is funding the semantic web technologies, notably DAML, in hopes of getting better AI -- this will be needed for better drones and autonomous agents, not to mention scanning for terrorist activities...

There's just no question involved. The military will do things no other organization would even think of doing.

More interesting constraint sets

[theonetruekeebler]

There is a certain sense of urgency in solving military problems. In that domain, requirements documents imply "...or many people will die". In a corporate environment, the requirements imply "...or we'll have to keep doing it the old way".

Then add to that "...and if it screws up even the slightest bit, under any circumstance, in such a way as to so much as hurt somebody's feelings, we're screwed." In that regard, the military provides a hell of a test bed for high-risk, high-concept toys, well away from the prying eyes of trial lawyers. Adaptive cruise control probably could not have been developed in a liability-conscious environment like, well, the real world. Without years in the hands of testers who knew enough about personal responsibility to be entrusted with extremely fickle multimillion dollar jets, your ludicrous SUV would be that much harder to drive inattentively. A decade keeping jet fighters about a meter from each other at supersonic speeds refined the product to the point it could be implemented in an environment that, while far more mundane, is far more expensive to fuck up in.