Keynote Address
Supercomputing '95

Wm. A. Wulf
University of Virginia



Thank you very much Sid. It's a singular honor to be invited to keynote a conference as prestigious and important as this one. Sid and I didn't talk about what our remarks would be before we started here, so I was sitting in the back there listening. One of the things that he said is also something that occurred to me when I first started to think about what I should say today, and that is the etymology of the word "supercomputing." If memory serves me right the prefix "super" got appended to a whole bunch of things back in the 70's; and I think it more or less coincides with the release of the first Superman movie. We got "Superbowl", "Super Tuesday" and a whole bunch of super thises and thats.

We don't have quite the same connotations to that word now that we had in those days. Certainly the notion of being big and being fast and powerful and that sort of thing is there, but I think in the 70s it also meant something about specialness, uniqueness, difference, apartness. Now, while "mainstream" isn't quite the right word to apply to supercomputing, it sure is getting close.

Let me point out some ways in which I think that's true. First of all its evident in the technology. We are now using commodity parts to build high performance computers, and simultaneously commodity parts are using ideas that were pioneered in very high performance supercomputers.

I am particularly heartened by another indication of the mainstreaming of the supercomputing, and that's the almost universal recognition of computational science. It is common now for people to refer to computing as the "third modality" of scientific investigation. Twenty years ago you really were an oddball if you were a physicist who computed. That's not true anymore, its universally recognized and respected.

Another evidence of the mainstreaming is the recognition that people who happen to use supercomputers also happen to need a whole spectrum of other things. The blue ribbon panel that Lou Branscomb chaired, talked about the "pyramid" of computing power as being essential to what computational scientists do.

I looked at your advanced program, and I said to myself "Holy Schomley, this is the conference teraflops or die?". You've got tutorials on virtual reality systems, on principles of color, on teaching technology for distance learning, and on ATM networks. You have two, count 'em two panels on community networking. My own University's demonstration is on a system called Legion, which is middle-ware. It's going to be running a supercomputing application but the truth of the matter is a few years ago this only would have been acceptable at a computer science conference.

Continuing in this vein of the mainstreaming of supercomputing - this is probably less evident to you than it is to me- but I spent alot of energy over the last five years looking at the use of information technology and high performance computing in humanistic scholarship. I am, in fact, at this point convinced that high performance computing is going to have more impact on scholarship of the humanities than it is on science.over the next two decades. That's partly because they're starting from further back, and partly because their problems are harder. It's been absolute fascinating for me.

My last point about the mainstreaming of supercomputing doesn't really quite fit but I was driving to the office the other day and almost went off the road because I was laughing so hard. It was because I was listening to NPR and they were quoting John Malone of TCI at the cable television conference/trade show. He said "A couple of years ago we thought interactive video or video on demand was going to be the application. But no, its the Internet." Wow! That's not just mainstreaming high performance computing, but another recognition of what I think this conference testifies to. As I look the program and as I look at the attendees, the issues you're dealing with are the integration of high performance computing, supercomputing, into and with the whole society.

Now to say that you're mainstream, is not to say that you're less important. In fact, quite the opposite, I think the acceptance by the scientific community, modulo a few troglodytes, of computational science as the third modality of scientific investigation is one evidence of the importance of mainstreaming. From my perspective at the Computer Science and Telecommunications Board, over the last five years there is an increasing public recognition of the importance of information technology to all kinds of societal concerns. There has been a qualitative change in the nature of our studies. I am oversimplifying to some extent, but a lot of our early studies were focused on fairly narrow technical issues. As you know, the Computer Science and Telecommunications Board is part of the National Research Council, we are chartered by Congress to provide authoritative and unbiased advice on questions that government may have. Most of the questions we were getting, half a dozen years ago were fairly narrow technical ones. We still get those, but increasingly we get questions that have much more to do with the societal impact: issues- intellectual property law, issues of import/ export controls, etc. We are now doing a large, Congressionally mandated study on national cryptographic policy, for example.

It made the rest of the academy sit up and take notice when the first visit that Newt Gingrich ever made to the Academy building was to talk to the Computer Science and Telecommunications Board, largely at his request. He said, in essence, "You guys have a responsibility that you are not fulfilling. You have a better perspective on what the society is going to be like technologically than anybody else, but you are not popularizing it. You're not preparing the general population to understand the kinds of changes and the kinds of transformations that are going to go on in society." We're still trying to figure out what to do with his challenge. It wasn't CSTB that Gingrich was coming to talk to, of course; he really was trying to talk to the community involved in information technology and high performance computing.

I repeat, just so I don't lose it - the importance is not that Gingrich came to talk to us; it is rather the increasing recognition of society at large, and Congress as a representative of society, of the importance of what you all are doing. I'll have a little bit more to say later about the responsibilities that go along with that. But OK, if being mainstream doesn't mean you're less important, I must say its not an unalloyed joy either. Let me give you three or four examples.

One that all of us are feeling is the pressure for resources. We have created a demand, we have created a larger and larger number of people who want to use computational resources, just at the time that federal support is declining. I think that's a source of angst of all of us.

A second "unalloyed joy" that I get sucked into a little bit, is that there are more and more experts in congress and on the staffs. Ten years ago, I could say almost anything, and if they didn't believe me, at least they didn't challenge me. We now have alot more people whose little bit of knowledge is a dangerous thing.

The third part of the unalloyed joy tis our greater responsibility to inform public debate and assist in policy formulation. This conference, is a Time Machine conference. This is the conference where we can talk about, where we can see what the general population is going to talk about and see in a decade. And so we can see here what public policy is going to have to address in a decade. Or maybe sooner!

The hardest thing for me to do, because its so contrary to people's experience, is to explain that what people think of as absolutely unobtainable today is going to be on their desktop in a decade, and that when they think about issues of public policy, when they think about issues about funding and investment, they have to think in terms of the way its going to be. It's not a question of whether something will happen, it is a question of when it will happen, and for the most part, it's going to happen faster than people recognize.

Alright, let me talk about resources for a while. It's not a pretty picture, but I think we have to face reality, and I want you to face five realities. There may be more, and these may be only "partially true realities", but, never-the-less we need to talk about them.

Number one: the federal science and technology budget will decline. It will decline. They said I couldn't have overheads. And this is one of the places I would have, so create a picture in your mind with me, for a moment if you will. I want to divide the federal government into four categories: (1) payments to individuals - that means social security, medicare/medicaid, aid to dependent children, veterans benefits, etc.; (2) defence - the obvious thing; (3) interest on the national debt; and (4) everything else - everything in the department of labor, in the department of agriculture, in the department of justice, in the treasury department, in the energy department, in NASA, in the National Science Foundation, and everything else is in that fourth category. I want to look at the percentage of the federal budget that went to those categories in 1960, and in 1993, and I am going to round the numbers off for simplicity.

In 1960 payments to individuals represented 25% of the budget; in 1993 about 60% - about a two and a half times increase. Defence in 1960 was 55% of the budget, and is currently about 20% of the budget - decreased by a factor of two and a half. Interest on the national debt went from 7% to 15% - it roughly doubled. Everything else, almost everything you think of as a function of government, as a matter of fact, went from 15% to 8%. As any of the thoughtful representatives or their staffers acknowledge, the only way to balance the budget is to address the issue of payment to individuals. The current trends cannot be sustained. There is only 7% of the budget to take anything out of. And as at least one staffer said to me, who I respect, "If things like social security and medicare and medicare are on the table, then everything else must be too. You can't think of anything as being exempted. It's just a political reality." So the federal science and
technology budget will decline. That is reality number 1.

Reality number 2: the decreases will not be uniform for a whole host of reasons. Some of them rational, some of them not so rational. Some of them tied to the way the committee structures happen to be formed in Congress, some of them having to do with the historical accidents of which agencies funded which kind of research. If the National Sciences Foundation is cut, it has a larger proportional impact on mathematicians than if the NASA budget is hit, whereas space physicists would be hurt in a complementary way if NASA was hurt rather than NSF.

Unpleasant reality number 3: A lot of my colleagues would like to believe that it is simply an issue of informing people in Congress of the value that science and technology - of giving them examples of demonstrating the values that science has had in the past. I don't care whether you talk about Teflon or penicillin or whatever, there are at least many thoughtful people who do understand those things. As another staffer that I also respect greatly said, "Science is everybody's second priority on the Hill." By that they mean there's nobody who is going to get elected on the basis of their votes on science. Most people in Congress have warm and fuzzy feelings about science, but if it is a choice between a cut in science vs. programs that affect their constituencies, it is no contest.

And there are some philosophical questions that I think are honest questions too. Questions about whether even more might have been accomplished if we had proceeded with some other model of funding.

Unpleasant reality number 4: One does not hear it discussed quite so much anymore, but I do not think the debate between "strategic research" as Barbara Mikulsky, the Senator from Maryland talked about it, and "basic research" (which unfortunately, at times, has been referred to as curiosity research), has gone away. I think the need to show societal relevance for what we do has not changed one iota. The rhetoric has changed a little bit, but the underlying issues that caused that debate in the first place have not changed at all. And I think frankly its a wrongheaded and counter-productive argument. And I'll come back to that point in a minute.

Unpleasant reality number 5: I think we are going to have an unsettling period in which various agencies and institutions sort out what the future is going to look like. The set of responsibilities, activities that have fallen to the various agencies, to the various kinds of national laboratories, and to the universities, are in part an accident of history. And I think that we should not presume that they are the right way for things to be.

I have just painted a really gloomy picture. So what are we going to do about it?

Well, let me suggest four things that we can do. The first one is simply to get beyond the denial phase. Things are going to decline. Not everything is rosy. Change is tough. Its going to affect individuals, and I do not want to down play that at all. But, every change has in it opportunity. The challenge for us here is to ask what is the opportunity and exploit it. Because I think there are many.

The second thing we can do is be states people. This has been said many times, but I'll repeat it because I think its true. When faced with invaders, we in the sciences and engineering tend to circle the wagons and shoot inwards. We presume a zero sum game in which the only way in which our parochial interests, our budget, can be defended is by taking it away from somebody else in the science community. I think that a very limited view and not a necessary one. Yes, things are going to go down, but that is not the same a saying that the only thing do to protect each of our little feifdoms is to steal it away from the adjacent one.

In that regard, at the end of my talk I am going to shamelessly steal some remarks by Neil Lane, the director of NSF. In a talk at the University of Chicago just a couple of weeks ago Neil made a very important point. He noted that the rhetoric to date has tended to focus on a social contract which was forged by Vanover Bush and Harry Truman at the end of World War II in which, as it is usually characterized, the Feds would fund science, scientists would decide what got done, and from that would flow prosperity, defense, and public health. And so the rhetoric goes, we have come to the end of that era and things are going to radically change and be different in the future. Neil's point is, "Well, no not really, the fact is that over the last 10 or 15 years we have been progressively changing the nature of that contract. And in particular the way we have been changing it is away from the simple bilateral agreements between the individual investigator and the funding agency. To much more pluralistic, much more collaborative, much more partnering kinds of relationships not just between individual investigators and the funding agencies, but between those investigators and other institutions and so on". What better example of that than the I-way which is sitting in this building today, which i think by the way is akin to a national treasure. He talks about the wonderful synergy and interplay between science and technology that provides those magic bullets for economic success and societal benefit.

What can we do number 3. I see great strength in the partnering that has happened between computational scientists and computer scientists; between university, industry and government and I cannot think of a better demonstration, better evidence, better testimony to that partnering than this very conference. The way the exhibits have been pulled together demonstrates a degree of collaboration that I think would have been considered absolutely impossible a decade ago. And so with Neil I think one of the things we can do is be statespeople and continue on this path of evolving collaboration, expansion. In fact, I worry sometimes is that this group being so highly successful, will take that success and turn it into a closed system to fend off those folks from the humanities, for example. God, you wouldn't want them to take some of our resources!

The last thing that we can do is simply recognize that it is up to us to do something, there is no entitlement. Society does not have a responsibility to support science and technology. Now, everybody in this room believes it is in society's interest to support science and technology, but that does not mean they have to. It is our responsibility to take the message to them and convince them of what I believe is in their very best own interest. And we will not convince them by being aloof.

The other unfortunate property of folks in our community it that we have tended, at times, to look down a little bit on the non-scientific community. And that will not work. It is up to us - but boy, that is a challenge I really like. I think in many ways this community is a role model for what the rest of the science and technology community is trying to do, or would like to do, or thinks they might think about doing some time in the future. The very successful mutual interplay between computer scientists and computational scientists is just fantastic. It is very clear that we computer scientist have benefited enormously from having you guys as guinea pigs. And I would like to think that you have benefitted enormously by having us running around and building systems for you. I said that in a light-hearted way, but I am very serious about it. It seems to me that this is a a perfect example of how two committees can synergistically play off each other to build something which is vastly more than the whole.

Correspondingly the university industry/government relationships in this field are very rich, very productive. I don't know how many of you have seen the report that CSTB produced on the high performance computing initiative. This is the report which is usually referred to as the Brooks/ Sutherland report after its co-chairs. It has in it figure one which I absolutely love. Figure one depicts a time line of the horizontal axis and vertically has a number of technologies: operating systems, networking, graphics, so on. And for each one of those technologies, the figure details when it was being worked on in university laboratories, when it was being worked on in private sector laboratories, when product development was going on, and finally when it became a billion dollar industry. There are a lot of messages in that figure; one of them is it takes a long time. Fifteen years is not atypical for something to reach the stage of being a billion dollar industry.

But the other piece of the story that figure tells is that it is not "linear"; it is not universities producing something which is picked up by the private sector laboratories, which is then produced, which is then sold. Rather, it is a very complicated picture. In fact, my wife redraws it to simplify it because its such a complicated picture. But I think the complexity is part of the wonderful story that it tells. Things don't just start in the universities, they bounce back and forth between universities and the private sector. There are products on the market and then the idea bounces back to the university. Our field has done that very well, and I think it has done it a fairly unselfconscious way. There is a higher degree of respect, I think, between, university people and industry people than in other disciplines. That is a very positive fact. But on the other hand, there are some more things that we need to do.

Maybe the first one that I want to mention is that the relations between universities, industries and government I think sometimes is less than it could be because we do not understand each other's value systems and reward systems as well as we might. I have had the privilege of spending part of my career in industry, part in government, and most in the university. One of the things that I am frequently struck with and disappointed by is the degree to which people in each of those sectors deprecates the others. The fact is that, let me just deal with universities and industry for a moment, the value system in industry is very simple, it is the bottom line, dollars, it is easily measurable. The reward system in the universities is much closer to simply pure recognition. That is a wonderful distinction. The fact that they are different means that the in principle one can build totally win-win situations. Industry can make money; academics can be recognized for their contributions, and neither detracts from the other. But it does not always work that way, because
academics do not understand the need to bring products to market quickly in order to make money. The industry people do not always understand the need for academics to be recognized
whether by publication or a bunch of other things.

There are other things that we still need to work on a lot. I think industry needs to "make the case", for example. It is really incumbent on the wealth producers to acknowledge the contributions made by both universities and government in this complex stew that we have evolved. I want to throw kudos to CSPP in this regard, I think they have been very responsible, but industry at large has a deep responsibility to keeping this engine of productivity going.

Academics: we need to be alot more flexible in a number of ways. The one that hits closest to the belt, and so let me pick on that one, is tenure criteria. Right now it is very, very hard for someone who has spent most of their life in industry to get tenured in an American university. No matter what their intellectual contribution has been, no matter what their stature is, they typically do not carry the right kinds of punches in their ticket. It is very sad; we have cut students off from immeasurable wisdom and experience.

Another one that academe needs to cope with, that I think is particularly relevant to this community is recognizing interdisciplinary work. I think we all know that folks who work at the boundary between two or more disciplines tend to be viewed with a somewhat jaundiced eye by the more mainstream discipline folks.

Universities also need to get their acts straight with respect intellectual property.

The government to make some moves too. It needs to be much more agile. We were talking this morning at breakfast about the problems of the procurement system. I think that's a particularly good example.

I am sorry if I offend some of the government folks here, but it is important that they become less risk averse. Over the past 20 years as I have watched federal funding agencies evolve, the ability to take risk has declined dramatically.

Government also needs to be less prescriptive. I think it goes hand-in-hand with being risk averse, to try to specify programs in a way that satisfies some criteria within the beltway, but do not fully exploit the intellectual juices of the academic or industrial communities.

But most of all, of all the things that we need to do, that the most important is exploiting the strength that this community really has. And that is the synergistic interplay between folks from different communities.

I want to pop all the way back to talking about supercomputing. The other reaction that I had when Sid first asked me to give this talk was to look at the word supercomputing and pronounce it as really, really super computing. What would it take to get really super computing? Well, I am going to plug an idea here, because in fact, it almost is happening out on the exhibit floor. A few years ago we first had a workshop at the National Science Foundation, and then we produced a National Research Council report on the concept of "collaboratories". For those of you who have not had the opportunity to see these reports, the word collaboratory simply comes from taking the word collaboration and the word laboratory and co-joining them. The idea is really very simple, it is to build a system that supports all aspects of science even though the researchers are not collocated. When we ac do that, we'll have really super computing!

So, as I warned you, I am going to finish this now by shamelessly excerpting some material from Neil Lane's remarks at the University of Illinois a couple of weeks ago. I will tell you I am taking some of these comments out of context, but I think if I read them through they very much capture the flavor of what Neil was trying to say
"With major portions of the R&D structure threatened, which is cause for great concern and anxiety, I believe we can still utilize this time as one of opportunity and we should. Foremost, we must prosthesize the unconverted about the connectedness and interdependency of the nations R&D structure. And not carve off the piece that we find most to our liking. It seems to me that something very different from the original compact, referring to the one with Vanover Bush, has been emerging in practice over the last 10-15 years. While many of us have remained wedded to the old rhetoric, the initial compact of the Vanover Bush era has evolved, perhaps without out conscious attention. What has taken shape is more expansive and multi-faceted arrangement that encompasses the larger R&D enterprise in the nation, both public and private. This new arrangement is characterized by fewer and fewer initiatives, that are confined to just two partners. Instead there are continuously evolving collaborations to serve the national interest, collaborations which exhibit, among other things, fewer boundaries and vastly more flexibility. And so, as we continue to speak with reverence about the compact between science and the federal government, we are in fact, thinking of the past more than the present. I want to emphasize stridently that I am not suggesting that we lessen, certainly not eliminate the federal role in support of fundamental research, primarily in our universities. Indeed, I believe more support is easily justified as being in the national interest. Rather, I am suggesting that this new perspective is a way to re-characterize the roles and positions of all participants in the National Research and Development System, as that system has evolved and matured."

Thank you very much.