Rick Rashid: MSR Faculty Summit 2004
Aug. 02, 2004
A transcript of remarks made by Rick Rashid, Senior Vice President, Microsoft Research, during the MSR Faculty Summit, Aug. 2, 2004.

RICHARD RASHID: Good morning. In the interest of full disclosure, if I seem a little punchy this morning, or if I make inappropriate references to horses, I just spent this last weekend crewing for my wife. She was doing an endurance ride at Paulina Peak in Eastern Oregon. I didn't wind up getting a whole lot of sleep the last few days, so if something weird happens, that's a good enough reason. The good news is she got a Top 10 placement, so she was very excited about that. And, boy, is that tough work. It was like 95 degrees they're riding in, too.

What I want to do this morning is talk really about the relationship of Microsoft Research and academia, and I know many of you are new, haven't been at one of these faculty seminars before, so there will be a little bit of repetition for those of you who have been before. But what I want to talk about is really, I think, the way we think about the partnership, how we think about what we do working with the academic community.

Now, just to give you an update on Microsoft Research itself, we're now about 700 researchers. We're continuing to both grow and spin out organizations into the greater Microsoft, as time goes on. Our most recent spinout occurred actually last week when we announced that we were creating something called the Center for Software Excellence, by spinning out parts of what had been our Programmer Productivity Research Center. And CSE now is going to be responsible for all of the key tools within Microsoft. They're part now of the Windows organization, and that's another sort of successful spinout as we move forward.

As I said, we're continuing to grow. We'll probably add about 50 new researchers this year, growth by another 7 percent or so. And for those of you who don't know, we currently have five labs. Our biggest lab is actually right here. In fact, you'll see many of our researchers over the next couple of days, but we also have very large facilities in Cambridge, England, right on the University of Cambridge campus, and a great relationship with the University of Cambridge. Actually our second-largest lab now is in Beijing. We have about 170 researchers in Beijing. They're growing very, very fast there, and having a big impact in the research area. We also have two other labs in the U.S., a small group in San Francisco working for Jim Gray, and a group in Mountain View on our Mountain View campus. So, that's kind of the state of Microsoft Research.

Now, you shouldn't be surprised that we're continuing to grow, because Microsoft as a company is really focused on continuing to grow its R & D investment, and continuing to innovate through the work that we're doing. I just picked up this slide, this is sort of an historical slide, and it shows how much we've invested in overall R & D during the period from '00 to '03. When a lot of other people were cutting back, we were continuing to grow our R & D investment. And, as a result, we've also been able to improve our corporate results during that period of time. So, we view that as really an important part of how we think of ourselves.

Now, there are a lot of great people working at Microsoft Research. This is a list of some of the notables that we have, but there are many, many others. Someone who has just come onboard, we'll be talking about more next week at SiGraph is Kurt Akeley, one of the original SGI co-founders, and one of the creators of reality engine and Open GL. So, we're very excited to have Kurt on our team.

But we're not just a place for great people, we're also a place where great researchers can really make their careers. And one of the things that's been exciting for me personally is the extent to which now many of the young researchers that came into Microsoft Research a number of years ago are starting to have a big impact in the academic community, and within Microsoft. Roughly 10 of the corporate vice presidents are people that have worked for me in the role that I've been playing at Microsoft, and so a lot of graduates from Microsoft Research now are playing significant roles within the company, but also a lot of young researchers are playing significant roles within the broader computer science community. I'm really excited about that.

Our mission has never changed. Our mission has always been exactly the same, and as long as I'm around here, it's going to be exactly the same because I really don't want to change this slide. You guys know how much work it is to make up slides. I always like to use the same slides over and over again if I possibly can. This one, I think I created in 1992, so it's now working on its 12th year as a slide in my deck. Our key mission is exactly the same as yours, which is to move forward the state of the art in the field of computer science. It has nothing to do with Microsoft. It's all about the field of computer science, and really impacting the technologies in that field. And so that's our first mission.

When we are successful, then we can take the great ideas, and the technologies we create, and move them rapidly into our products, and that's really our second mission which is, when something works, you know, research has a lot of risk, when it works, work like the dickens, get it into the products, and get it out to people. And we certainly do that.

Ultimately the reason we're here, and the reason we work with you in the academic community is to make sure that Microsoft is still going to be here 10, 12, 15 years in the future. And while that always sounds like, oh, sure, you guys are going to be around for a long time, the reality of our business is that companies come and go. They come and go based on their ability to continue to innovate. And so it's important for us to continue to innovate and stay at the state of the art in all the areas that we do our work.

Now, the reason I think we've been successful building Microsoft Research as a research lab is because really I've done it using your models. The goal that I had coming to Microsoft back now almost 13 years ago was to create an industrial computer science research lab around the model of university research that I had when I was at Carnegie Mellon. So, we're really a lot today like Carnegie Mellon University was back in the mid-1980s in the way we think about ourselves, the way we organize, the way we do the work that we do. And, again, I think that's a testament to the value of that approach in being able to create great research.

We have very strong ties to universities, and I think the testament to that is the fact that even going back to the very earliest days of Microsoft Research, when we only had about 15 or 20 people, we were even then investing in universities, and over the years more than 15 percent of all the money that has come into Microsoft Research's budget has gone to universities in one form or the other, whether it's fellowships for students, whether it's curricula grants, whether it's research grants, lab grants, and what-have-you. So, that's always been an important part of the way we think of ourselves.

Now, I said, we're focused on moving the state of the art forward. I updated this slide a little bit to show some of the things that we've been doing in 2004. We have a huge number of papers at conferences, especially this summer going into the fall, the production of the organization is just tremendous. We've got a large number of papers at SIGGRAPH this year, at SigIR, SigMod, OSCI, and a number of other conferences. So, I'm just really excited about that, we've put a lot of emphasis on getting our research published, and getting it out the door, and working with the academic community. This is just a you're not really supposed to read this, you're supposed to just look at the slide and say, wow, that's a lot of stuff. These are some of the papers, we have 13 papers at SIGGRAPH next week, and so these are just some of the things that are coming up in that conference.

I mentioned our second mission is to move technologies into our products, and we work really hard to do that. And a lot of the things you see today at Microsoft have come out of Research. A lot of parts of Microsoft have come out of Research. What is now our Digital Media Division really started with a group I created in 1993 in Research, we built them into a product organization and spun them out in 1996. And they continue on today, more than 700 people in that organization. I was involved in one of the first e-commerce groups in the company, the work we've done in data mining and SQL all came out of Research, all of our speech and natural language products, the Tablet PC work, BOT, our Smart Personal Object work all came out of the Research group. So, really a substantial part of what is Microsoft today came out of the work that we've been doing in Research. And here are just some of the things that we've been doing just over the last 12 months or so in passing our products.

One of the things I'm very proud of, and as I said we just announced this spinout, is the impact we've been having on the way Microsoft does its software, the way it's actually implementing its products, and a lot of key new tools have gone into our product process over the last few years. I've just shown you a list up here of four things, PREfix and PREfast are holistic tools for doing software analysis, and basically doing model checking of very large pieces of software, meaning the entire Windows development environment, the entire Office development environment, and all of our other key products. And today, developers are required to put their software through these tools before they can check their code into our source trees.

SLAM and ESP are actually principal tools that prove the properties of large programs of hundreds of thousands or millions of lines. SLAM now externally is being used as the static driver verifier that's part of our DDK for Longhorn. ESP is tool we've been using internally to do sound analysis of our interfaces and many other parts of our Windows environment. Again, I'm really excited about that, and the fact now that that group has literally spun into the Windows division is an indication of how important what we've created within Research is to the future of Microsoft.

Now, we would not be successful as an organization if it weren't for the fact that we've had very, very strong ties to the academic community from the beginning of this organization. From the outset, our governance has included academic advisors for each of our major labs. So, in the U.S. we have a technical advisory board that has been now running for nearly 13 years. We have a great board of people in Europe, really taken from top institutions all across Europe, and likewise in China. And there, in fact, we draw broadly from Asia board members for our Microsoft Research Asia lab. So, that's an important part of the process.

But we've also had a huge input from the academic community in the work that we do. If you look at papers we publish, a lot of then are joint papers published with professors that are either visitors here, or that we've had collaborations with externally, interns that have worked with us. Just this year in Redmond alone, we have 184 Ph.D. interns. That's just a huge number. There's roughly 800 or 900 Ph.D.s currently produced each year in the U.S., so a significant fraction of that cohort each year has spent at least a year as an intern at Microsoft Research. And around the world that number is close to 300. So, again, it's a very significant part of what we do. And, in fact, in Asia, we literally run an educational program. We have the rights from the Chinese Government to grant degrees in computer science. We have 250 students in that program receiving a post-doctorate degree in computer science from us. So, again, that's a significant part of the way we think about ourselves, and the way that we do our work.

We also have been involved in doing a lot of symposia, typically working jointly with universities, the University of Washington, CMU and others; as we attack new areas, we try to bring researchers together to attack new areas. This year we had a great Mesh Networking Peer-to-Peer Conference, actually about a month ago now, up here in the Redmond area. Social computing conference is very successful back in April. Spam conferences coming up, down in the Bay Area. Again, really trying spam and e-mail work in general is what we're looking at there. Really trying to say, bring the community of people, often cross-sections of industry and academia together to attack critical problems or to look at new opportunities and new areas. And that's, again, something that we think is particularly important.

Because of all this, I really view a key part of our mission as an organization is accelerating science and education. So, a lot of what we're doing in our university relations program is really focusing on how do we accelerate the work in science, and not just in computer science, in some cases how do we take computer science and help accelerate other fields as well, and how do we have an impact on education. Without the educational process being successful, we are not going to be able to drive the technologies in our field that are going to allow us to be successful in the future. And by us, I don't just mean Microsoft, I mean the field. If we don't create new value, if we're not moving the state of the art forward constantly, you know, our relevance to society goes down. So it's important to make those investments.

We've literally supported hundreds of research grants over the years to universities, really trying to accelerate science. Just in the last three years alone, we've invested more than $100 million in universities just in North America. So, I think you get a sense of the impact that we're trying to have. And I'll mention later one some of the collaborative research that we're doing as well as the attempt, as time has gone on, we've realized that opportunities to create new kinds of platforms to support academic research or support education, and so many of our investments, especially new investments going forward, are looking at how do we create long-term platform value, in effect, in that world.

If you think of Microsoft as a company, we're sort of a platform company. Can we take some of that thinking and also use it to move forward science education? An example of this is the work Jim Gray has been doing on SkyServer. And I know some of you know this work; some of you may not, I'll quickly go through it. Back in 1998, Jim and his team did something called TerraServer, and we put up initially a terabyte, and now it's close to 20 terabytes of data on the surface. And I'll point out, by the way, TerraServer is still running. It was originally a system about a quarter of the size of this room with all the disk drives in it. Now the system is dramatically larger, supports more users, and it's actually just two small racks about this high. And it went from being about $2 million worth of hardware to being about $70,000 worth of hardware. So that continues. We now have half-meter resolution data, so you can actually see people now in the images. You can't really tell who they are, but as we get better data from USGS maybe we'll be able to do that too.

SkyServer is really an attempt to take that same technology concept and bring it to the astronomy community, and the notion here is can we federate the great data from the great telescopes of the world into a global federated database that allows scientists to have a 24-hour view of the best astronomy data and be able to do data mining against that. So that's what the SkyServer work has been about. SkyServer.spss.org, if you want to go out and look at it. This is just what the Web site looks like. The great part about this is that it's set up both for laypeople, the layperson can come in look at this and say, wow, these are really pretty pictures. And have a great time, and learn a little bit about astronomy at the same time. A scientist can take that same pretty picture, click on it, and now get all the articles written about that phenomenon, as well as get the spectral data in a broad category of information all in standard schema that they can do data mining against.

In fact, new phenomena now have been discovered by being able to do data mining against this information in a standard way. So, it's no longer all dusty decks of code that nobody knows what it does, and databases no one knows how they're formatted. It's now all done in SQL. It's all available, you can datamine against it. With three lines of SQL, you can find every quasar in the Sloan Digital Sky Survey. Just a huge improvement in what these scientists have been able to do to get access to their information. And it's changing the way they think about sharing information with each other.

It's also a great educational resource. If you go into the same Web site, and there's about 150 hours of educational material. Here's an example of a project that's online that a student can complete, where they can actually rediscover Hubble's constant, using the best data from current telescopes. They have the exact access to the same data that any modern astronomer would have, but now they can do their private projects with it, or teachers can use this material to create their own curriculum. So we're really excited about that. By the way, Curtis Wong, another one of our researchers, designed the educational component of that Web site.

Another area we've been looking at is, again, taking computer science techniques and reaching out to other communities to help accelerate sciences in computational biology. And there we recently announced some work we've been doing with the University of Washington, analyzing AIDS vaccine candidates, using machine-learning techniques. And we're now in the process of evaluating the notion of building a Web service to actually do vaccine candidate analysis. And we think that's a real opportunity to help that community move forward. And the people in the community are very excited about that.

Really, again, service, thinking of it a little bit like a SkyServer notion, of can you build some basic services that help accelerate work in a particular discipline. And we're looking at a lot of other things, I'll just throw this up there really quick, whether it's working with genetic diagnosis, RNA inference folding, and so forth.

We're also looking at not just doing things ourselves, but collaborating with universities. Again, I don't really expect you to read much of this, butiCampus is a program we started now a little bit more than four years ago. It's just completing its fourth year. And it's had a big impact at MIT, over 300 MIT researchers have been involved in the iCampus research program. Over 100 courses have been taught, built around work that's been doing with iCampus. It's really touched more than 5,000 students, 30 masters thesis have been done, supported by this work. So that's been a great program. And there's an opportunity later to see demos from the iCampus work, and to learn more about it.

Another effort that we've been involved with is the Cornell Theory Center. Again, the idea here was, can you take some of the value that's been created in the commercial software world, and use it in the high performance computing community. There have been, again, a number of successes that have come out of that, really looking at things like protein folding, and structural biology, computational genetics, and so forth. And that, interestingly, has had an impact back on Microsoft.

Now, Microsoft has a product team in the high performance computing area, and really many of the influences on that have come out of the work that was done with the Cornell Theory Center.

Some more recent activity, some of you may remember, Bill actually did a demo of DLN from the University of Maryland here a couple of years ago at the faculty summit, and that was a very successful project that we had going on with the University of Maryland. And now we're actually moving forward and creating something with UMB called the Microsoft Center for Interactive Design and Visualization. So that will now be a new center there, three-year funding from us, to really attack many key problems in interaction design and visualization.

In terms of new initiatives, I'm going to talk about three things, and effectively announce three things that we're doing. Work we're doing on development of new talent, educational development, curriculum development, really improving the educational process, and as I said, helping to build platforms, and accelerate science. So let me talk about each of these three.

One of the things that you'll be hearing more about over the next few days is we're announcing today a trustworthy computing curriculum RFP. And the goal there is to dedicate $1 million over the next two years to help develop curriculum in Trustworthy Computing, and really improve the state of the art, in terms of education for students in CS programs, really attacking broadly the problems of trustworthy computing. This isn't just about computer science, it's also about business and law. And we'll be accepting proposals in business and law, as well as in computer science as part of this program.

So we're really excited about this, there hasn't been really a well defined curricula in this space. And what we're hoping is by helping to seed new work in this space, not research, but curricula work in this space, we'll be able to accelerate the rate at which great new courses, and new materials will be available to people, to teach trustworthy computing.

Another announcement we're making today is that we'll be doing an RFP around the Phoenix technology. Phoenix is a next-generation proprietary technology that we're going to be sharing broadly with the academic community over time. Phoenix is based around many of the key things that I mentioned, that have come out of our PPRC, and are now part of what we call our Center for Software Excellence.

These technologies allow, we think, for researchers to really advance the state of the art in code generation, optimization, programming analysis, binary transformation, and software correctness. So we're really excited about this, as well. This is both about putting code into the community, to help make things happen, but it's also about providing funding to really move forward the state of the art, accelerate the state of the art in the whole area of software correctness, and software generation.

Finally, the last thing I'll mention is really focused on the future. I'm excited about this partly because it's a very forward looking, future view of things. You have to think about, when people ask me about technology transfer, when they ask me about research, and the research agenda, almost always what I'll say is, it's really not about technology, it's about people. Having the right people, creating the right people, over time, and really accelerating the rate at which great people go into the field of computer science.

So one of the things we're beginning, and announcing it, is to create a new faculty award that really tries to support some of the key new leaders, and thought leaders in the field of computer science. And to some extent, incent more young people to go into the field of computer science. Again, there will be a lot more information about these programs as they go forward. But, I think the key things I'm trying to leave you with here, is what's important to us in working with you is accelerating science, partly by creating curricula that improves the state of knowledge of the students, bringing great people into the field of computer science, doing the development side of things, and building platforms, and creating technology with you that will move our field forward, and really have an impact out into the future.

So I think the next couple of days are going to be pretty exciting. There's going to be a lot of other stuff that's going to be talked about over the next couple of days, other programs, I'm just hitting the highlights here in what I'm saying. And with that, I'm going to be done, and I would like to introduce someone who, honestly, he'd be better off introducing me.

Many of you have seen Bill speak here before; if not, you've probably seen him on TV. So without further ado, Bill Gates, our chief software architect and chairman. (Applause.)

Read More: