![]()
![]()
![]()
Click Multimedia PI.HTM Click Multimedia PI.HTML
Dar Click para Web Page Remarks by Bill Gates, Chairman and Chief Software Architect, Microsoft Corporation Cornell University, Ithica, New York February 25, 2004 BILL GATES: Thank you. It*s great to be here at Cornell. Cornell has made some fantastic contributions in many, many different areas, particularly engineering and computer science, have a lot of cutting-edge laboratories, and we feel privileged that Microsoft has a strong relationship with a lot of the activity going on there. Certainly, Microsoft has benefited fantastically from hiring graduates from the schools here. Steve Sinofsky, who runs the Office Software Group at Microsoft, a very key person who was mentioned, and the incredible way that his experience here helped get us really behind the focus on doing super Internet software. Another graduate, Mike Nash, runs our security activities. You can imagine that*s a reasonably important job, and one that keeps him very busy. But we have people from many parts of the company, the head of our human resources group, and many, many other people, and so we*re pleased at the connection that we have. What I want to share today is some of my excitement about the impact of breakthroughs in computer science, the impact that will change the way we do business, change the way we learn, and change the way we communicate and entertain ourselves. In many ways, this is a field that people are now underestimating. Although in the late *90s and bubble period, there was a lot of hype and a lot of talk about things happening overnight. In fact, what will happen in this decade goes beyond even those wild dreams, and it*s happening because of the advances, infrastructure changes -- primarily in software -- that we and others have been working very hard on. The impact will also be on how science is done in many other areas. The tools of computer sciences, the models in computer science, will really change. And we got a slight glimpse during the 1990s of the impact of the extra productivity that came out of a modest set of software advances. So here, we*ll see a much greater productivity benefit that won*t just be impacting the United States, that will actually change the global market, and in fact, increase the effectiveness of global competition in delivering great products and services to every one around the world. My history, as was said, has one unusual thing, which is that I am a college dropout. I*m here today advocating staying in college, finishing up, and really getting the great background that you can only get by finishing your school. The thing that convinced me that the time was right to start Microsoft, even though it wasn*t super timely from my parent*s point of view, was the arrival of a microprocessor-based system. It was a very different than computers that had been up to that point. Computers before had been tools of large organizations, and people feared them because they would send out these funny bills, and you couldn*t get them to correct their information. People even talked about folding and mutilating and stapling those computer cards that you*d get as part of your bill just to fight back against what the computer was. So, this idea that the PC would become the most empowering tool we*ve ever created -- the best tool for creativity, communications, publishing, and really alter every one of those activities -- wasn*t widely anticipated. My friend Paul Allen, who was the co-founder of Microsoft, saw the 1971 8028 microprocessor from Intel, and by extrapolating out the Moore*s Law prediction of exponential improvement, he saw that we*d have something very different. And he enlisted me and said, *Hey, let*s start a company to write the software for this machine.* Well, that idea was a very strange one, because, of course, all the software at that point was written by the hardware companies. All the machines were incompatible -- the IBM machines, the digital equipment machines, Univac, NCR -- utterly different, and the software was written inside those companies, the operating systems and the tools. And so, our idea was that we*d make all these machines compatible, we*d provide the software layer that ran on top of those machines that would allow people to write applications to sell a volume of applications proportional to all the machines in the marketplace. And that would create a virtuous cycle, the more applications were there, the more people would buy, that would increase the volume, which would reduce the prices, drop the cost, and draw more people in. So starting with that 1970s kit machine, a cycle was begun. The early machines were not very capable. That Altair was a kit. All it did was flash the lights, there was no disk. The most advanced version had 8k of memory, 8,000 bytes of memory. And one of my great programming feats was writing a basic interpreter that could run in an 8k byte environment, including floating point and string management, and storing people*s programs. So you can imagine in product review meetings now, on days when people come into me and say a piece of software is 8 megabytes, or 20 megabytes in size, I kind of shake my head and say, *Wait a minute, how did it all get so big?* And maybe it shouldn*t be so big, that*s definitely my initial reaction. But, the scaling effect of this additional power led to the second generation, which was the Apple II, TRS-80, Commodore 64, all of those running Microsoft Basic built into the machine, and then in 1981 to the IBM PC. In some ways, we think of that as a pretty limited machine today, but in its time, increasing the memory capability, that was pretty exciting. And as that became more and more powerful, that led us to graphical interface. That was something that Xerox had really played around with in their labs, the Palo Alto research labs, but it was Apple and Microsoft that took and created commercial products that popularized that approach. It*s probably not easy to remember that it was very controversial at the time, people thought it was too hard to write the programs, it was too slow, and it was just a frilly thing. Good old characters, mono-based characters were all that people need, or needed, even lower case was considered sort of a lower thing, that "real" programmers wouldn*t ever engage in using that, and we still see that a little bit in some of the old time programmers comments, or lack of comments, the things they do. Then Windows came along. It became a phenomenon, and that was the start of our building productivity applications -- Word, Excel, PowerPoint -- and that*s become a really phenomenal business, as people are able to interchange documents, and as those products become richer and richer. The late *90s brought the wild period where the Internet, based on work in universities including Cornell and many others, exploded onto the scene. And people realized every computer on the planet would be connected. And that kind of really opened up people*s eyes to what was possible. And in many ways, completing those predictions, that will happen this decade, even though people are underestimating how that*s going to come together. Certainly, if we look at the PC today, as great as it is, it*s clear there*s a lot more to do. It*s not that easy to use. You have to learn various commands in different programs, even for dealing with common ideas like lists of things, lists of mail, lists of files, lists of music, all very different. Way to many verbs, way too much user interface, way too difficult to move your information around, to search it, to have it on the different devices. Way too complicated, it*s all aggregated there together, and you can*t connect it up to any display or any speaker that you want to. And so, I*d say in some ways, we*re maybe a third of the way to achieving the original vision that Microsoft had, which was a vision about empowering everyone with this tool. Now, one of the great helps we have in moving forward is the continued improvement at the hardware level. For the microprocessor, Moore*s Law, which has stayed true from the start of Microsoft to today, will certainly stay true for another 10 or 15 years. The one question there is whether we can map the increased transistors directly into performance, because the idea of parallelization is one of the tough problems in computer science, and we will need parallel techniques to be able to continue to map the increase in transistors to an increase in performance. We won*t be able to use just a purely brute force approach. Now, there are several areas that are improving even faster than the microprocessor. This is really hard to believe, but the capacity of the disks, the magnetic storage, on these systems doubles more like every 15 to 18 months, and what that means is already you can type an entire lifetime and not fill up that disk. But as the disks improve, the rest of this decade you can take all the movies you want to watch, all the photos you*ve ever taken, and store those as well on something that*s very, very low cost. So storage is no limitation in terms of getting to movies, and photos, and very, very rich things. The screen is another key element. We need large, very high-resolution screens. If we look at the impedance today between the world of paper, where people are still reading periodicals and taking notes and printing things out for meetings, one of the reasons that can*t move over to the digital environment is because the readability just isn*t as good. The fact that it*s often in a fixed position, the text doesn*t look as good, that*s a great inhibitor. And we*re committed to making it so that reading is far better off of the screen, where you can search and annotate, you can share your thoughts with people, the information is very much up to date. But we have to match some of these characteristics that are making people live in a split world. Certainly, the Tablet computing device that comes with a pen and a digitizer is a great step in that direction. The readability there, the idea of annotating things, all build in now is moving us from just having a keyboard to now having a keyboard and a pen as ways of getting information into these systems. User interface experts need to think about these big screens and how we can take advantage of those, the way we did Window management in the past definitely was influenced by having displays that were, say, 17 inches, or 18 inches or less, today I have three 20-inch screens, and although that*s a fairly premium price configuration today, that will come down to $700 or $800 in the years ahead, and be a very typical thing that can drive productivity. The graphics processors in these systems are really phenomenal, and we*re using them for entertainment software, but mapping those into things like seeing a representation of a bookstore, or a laid-out set of documents in a rich 3D way, those are techniques yet to be invented. Clearly, 3D will move into the mainstream interface, because the power is there to do that. The wireless connections are exploding. Wi-fi is one of those technologies that was really underestimated, but now it*s becoming pervasive, in the business setting, university setting, home setting. That means that the information can be moved around. And that will be complemented at small distances with ultra-wide bands that have unbelievable bandwidth to connect up to your screen or your disk without it being part of the same device. WiMAX is a standard that*s at an earlier stage than Wi-fi, but it*s a long-distance standard. It will take the cost of broadband connections and make them low enough that they*ll become practical even in rural areas in developing countries, where those connections are not very prevalent today. We see a world of many, many devices, and this I think fits in with some of the vision and exploration going on at the university here. We see the wall-sized device, the desktop, the Tablet, pocket-sized device, even a wrist-sized device all working together. And as soon as you indicate a preference about whether you want to see certain sports scores, or be told if a flight arrival changes, you should be notified on whatever device you happen to have with you. The system across all the devices should have a sense of what you*re doing -- that is, the context -- as well as a sense of what you care about. Is this e-mail worth alerting you about? Is this phone call something that should be scheduled to happen right after your meeting is done? And software is working for you to make all of that work. I mentioned moving from the pocket-sized all the way down to the wrist, and what I*m wearing is a product called the SPOT watch. We just came out with this about a month ago, and it*s got a special microprocessor and radio set of chips that we designed built-in. The microprocessor here is ten times more powerful than the original IBM PC. It runs at 30 megahertz. It*s got 640k of memory standard, so that*s ten times as much as the original IBM PC. And so, we download arbitrary programs that we can update just using the CLR byte code for things like watching sports of various kinds, looking at stock information, looking at weather, messages, calendar; anything you*re interested in can show up here, whether it*s as simple as just a customized personalized watch face all the way up to business information that might be interesting. And so, having information at a glance that you don*t need to get anything out of your pocket -- you don*t need to make a point-to-point connection, it*s just a broadcast network -- that always has the information available, we see that as just fitting into the hierarchy of a world with many, many devices that are working on your behalf and delivering key scenarios. Now, what holds back having all this great hardware result in fantastic things or result in great productivity? The answer is it*s all about software. I*m biased, but software is where the action is. Good software will fulfill these dreams, and we don*t have everything we need there. Just think of various domains. Think of what we call an information worker, somebody who as part of their job has to organize things, whether it*s a salesperson, a purchasing person, a product design person. The vast majority of the U.S. economy are people who do some type of information work. And yet, today the way they track what*s going on, the insights they get into their customers* attitudes, the way they can explore quality metrics, the way they even can look at basic things like sales trends and data mine through those by region and product type, pricing approach, they are literally starved for information. The data they get today is not good at all. Whatever data they get, they get on a piece of paper, and if they look at the number on there and say that*s bigger than I expected, it*s way too hard to dive in and say, *OK, why is that different?* If it*s up on a screen, they ought to be able to pivot through it and have analysis -- business intelligence software -- help them look at what the explanation is for what they*re looking at. So we*ve got to make it live, we*ve got to make it at a much higher level in terms of how they talk about it, model it and share it. Just look at meetings. Meetings are a huge part of people*s schedule at work and yet most people would tell you half the time in these meetings is not well spent. It*s information that doesn*t matter to them, that could have been sent out in advance, that doesn*t get followed up on. And by using software to facilitate the meeting, record the meeting, let people at a distance participate in that meeting, there*s a lot we can do to make things far, far more effective. The world of collaboration is just at its beginning. If we look at software customization, most businesses take application software and write lots of lines of code to customize it to their needs. That*s very expensive, and it*s the wrong level of abstraction. Instead, they should be taking a visual business process and relaying that out with the events and things that really are different for them, versus the other businesses in that industry. And by doing it that way, as there are improvements in the basic underlying modules, you don*t get a conflict between the customization work and the base improvements that take place. So, we*re at the wrong abstraction level, and modeling tools -- software modeling tools -- are what will close the gap there. Even the basic process of buying and selling is very inefficient today. Finding anyone who might sell you something, checking their reputation, checking the status of the order, dealing with something complex like when you get an invoice that*s wrong but you want to keep the goods if they*ll adjust the price, and not getting tied up in the mismatch between the ad hoc e-mail and phone calls versus your back-end software and that other company*s back-end software that doesn*t really understand those exceptions -- modeling these things very explicitly lets us track them, lets us manage people*s time in an effective way, and really puts them in charge and makes it work the way that it should. Communications is another great example where things are clearly inefficient. Why do we have phone numbers? Why do we have many phone numbers? Why do we get phone calls that interrupt us when we*re not interested? Why do we get e-mail that wastes our time? Our time is a valuable resource. Now, some of this e-mail that we get is actually almost humorous. I*ve got a few examples of some that I*ve gotten recently. This one here is pretty exciting. (Laughter.) It turns out if you get out of debt you get to meet people that are really friendly to you, looks good. Another one I*ve gotten looks like it might be more targeted. (Laughter, applause.) And I haven*t responded to this but I like that look with the diploma in hand. And finally the one that probably I am going to have to follow up on -- (laughter) -- is this legal thing. Whoever sent me that has got very good targeting software. It*s something that would be very, very timely. So spam is wasting our time. It is a very serious problem. It can even be used to sort of fool people into doing something they shouldn*t do, ignoring messages from other people, so we*ve got to make e-mail authenticatable in terms of exactly who it came from. And we have to give people control methods so that only the e-mail that comes from people they suggest gets passed through, and e-mail from strangers is subject to various proof techniques that make sure that it*s appropriate and what they*re interested in. We have to take the mismatch of all these communications modalities -- instant messaging, e-mail, phone calls, wireless versus wired, blogs, the blog indexes you get -- and bring those together in a simpler way. You shouldn*t have to join a game network and a social network and set up your e-mail and set up your personal Web site as very, very disparate things. And so, this is a hot area, a lot of advances in communication have a very profound effect in every realm of activity. For consumers, the move towards digital is really under way in big numbers. Digital photography is more popular now than film-based photography. And the fact that we don*t just think about it as photos, we can think about taking a set of photos and having the computer, for example, make sure that it picks the people who are smiling in each shot and create a collage so you don*t have to make a trade-off of who looks bad in this one or who looks good in this one. That should be automatic. Recording audio so you can really talk about how you felt about the event -- we have a thing called Photo Story that starts to say that you shouldn*t take motion video and stills and audio, and how you pull those together and think of those are completely separate. We need some unification there. The idea that your memories can be tracked for you and easy to navigate -- I think people would find that of immense value. One of the people at Microsoft Research actually has a little camera-like device that she carries around throughout her work day, and it*s noticing where she goes to different places or when people are laughing or talking loud, and it just passively -- without being noticed -- takes photos. So over the course of the day there will be something like a hundred photos, and it would be great to have software that can kind of sort out which ones are important, take the GPS data and the time data that*s associated with those, take the information off of her digital calendar and do the annotations to make that work very well. From that calendar, between what*s there and the state-of-the-art in face recognition, even being able to point out who*s who, and knowing what*s in the photo becomes very straightforward. Now, with all this proliferation of media types, we have to do better visualization, better interfaces. I*ll just quickly show you two prototypes that are being played around with in Microsoft Research that suggest that this is something that can be handled and made quite attractive. The first one is very simple; it*s called Media 3D. And it*s … say I*m interested in looking at different movies. And so, what we do is we take a movie, we put it at the center, and then on the outside we take the actors, the director and we put films that relate to them. So if I go up here, I can rotate through and see what things Ridley Scott did, and if I see one of those I pick that, bring that to the center and then it goes out to the database, brings up all the different clips and you can see now that Michael Douglas was in this and so we can see the movies about him that he is in, we can see Andy Garcia and various people. And this could also be annotated with whatever movie reviewers you trust, whatever top lists it*s on, what your friends whose advice you value thought about the thing, bring it together to a very rich and kind of easy to navigate interface. The other prototype I*ll show you real quickly is called the Media Browser, and this is more about photos and film clips that I mentioned. And so what it will do is it will load in -- in this case, I*ve got more than a thousand different images here -- and it*s putting a lot of those up on the screen using kind of miniature format. I can see I*ve got quite a variety here. Some of these are actually film clips, so if I double-click on those, then it will actually go and play the little film clip that I have there. So all of these are essentially in a database, but you don*t really want a database-like interface for navigating these things. You want to be able to select them easily, being able to put keywords on them easily. Well, let*s go look at the photos that have been tagged with *Thanksgiving.* Now, it*s likely that these were all taken in a similar time period and a similar area, and so it was easy to group them. I can say which of these photos have faces in them, just by hovering over that and it selects those. I can say which of them were taken indoors, which of them were taken outdoors. And the software recognition that*s helping with this is not perfect today, but it*s very good, and it makes selection and navigation a lot easier. I can even say, *OK, which photos are similar?*, so I*ll take this bridge photo and select that and then I*ll say, *OK, relax the constraint and say what is like that,* and as I relax the constraint on similarity, I get more and more photos that look like it, so I can select those and tag those in any way that might make sense. If I go back to the top where we had all those different photos, this is also a place that we*re playing around with 3D and saying, *OK, how could that help us group things in an interesting way?* When I have these stacked, of course, I can still go through them. I can also select a set and put one of these tags on or I can actually say I want even more groups to show up there, so I just take this slider bar and change that, and then it will regroup things with a little bit of more granularity on the X axis if that*s how I want to do those groupings. And so I think you get a sense here that there*s a lot we can do that*s very different than the way that we*re navigating the media today. One of the things that makes this compelling is in the world of new devices, people will be in control of when they want to watch video and how they want to watch their music. We*re moving away from the world where you put a CD in and listen to those tracks and you have to go get that physically or where you watch TV shows when they*re scheduled to be watched. This device here that I*m holding is called a Portable Media Center. It will come out from Creative, Samsung, quite a few different people this fall, and it*s got a nice color LCD and a 40-gig disk. And if you connect it up to a USB, automatically the TV shows you record, your film clips, your photos are brought down to this and, of course, music as well. So it*s kind of a superset of portable music players, but it*s really getting the movie companies and the music companies to think, *Hey, we really have to do a better job of making it easy to license software and have flexible rights so I can use it on many different players.* And so these, I think, will become very pervasive. Whether it*s a kid watching a movie or somebody on the plane, the fact that you have exactly what you want whenever you want -- it really puts you in control in a different way. It*s very interesting in terms of licensing models and advertising models the effect this is going to have on the media world, but things are moving pretty quickly because that*s what users are demanding. So I said that software advances are the key to this. We show our optimism and commitment to driving that forward by our R&D investment. We*re spending $6.8 billion on R&D this year. That*s substantially the largest of any technology company. And it*s pretty focused, in a way. It*s not on physics, it*s not on biology, it*s focused on software, and really focused on a single, unified architecture around Windows and XML and Web services in a way that has a coherency and doesn*t treat things like management or security as being off on the side. They are at the core of how we do this design. A big part of our R&D, a very important part, is our pure research group. It*s got sites in Cambridge, at our headquarters outside Seattle and over in Beijing in China. And that*s the group that tends to work most closely with the university here. In fact, we have some very notable relationships. The Cornell Theory Center is actually one of the biggest things we*ve supported, and it*s been fantastic for us to understand how Windows can be used in high-performance computing, look at the different applications that are emerging there, make sure our tools are very good for that, and so that*s been extremely fruitful. We*ve got Windows CE, which is our sort of mini Windows being used in some robots and vehicles to help out there, and we*re very excited about the work coming out of that. Our biggest R&D area is security. We call this broadly Trustworthy Computing. There is no doubt that the dreams of commerce and media and great things around the Internet and PCs, really only one thing could stand in the way of that happening, and that is if people perceive that their data and the reliability and privacy around these things just aren*t well understood. So that*s why for the industry, this has become very much a top issue. It reaches down into the very Internet protocols and how they were originally designed. There was a certain robustness against parts of the network being blown up or out of service, but not even there with the right type of guarantees. There was no design for authentication, for knowing that if somebody was malicious on the network that you could eliminate that traffic and not be fooled by the things they were doing. In order to solve this problem, there is innovation at many levels. Things like updating software and firewalls are very much near-term solutions, and just the great progress taking place there will make a huge difference in changing these things. But over time, the very way that we write the software, the fact that we can verify properties of the software, that we write essentially in a higher level language, in a more tight language that has contractual guarantees between the modules, allows us to prove things out piece by piece. Proving that software is correct was something that was being played around with 25 years ago when I left academic computer science. And I was a little worried that I*d leave and just overnight they*d have some breakthrough in it. That didn*t happen. In fact, it*s only really now, with the collaboration between Microsoft Research and a number of universities, that doing that for large bodies of code appears to be a very practical thing, and a tool that we*ll use not only for our software but we*ll provide those tools to our customers as well. Getting up from hundreds of lines of code to millions of lines of code is a very tough problem, but on things like device drivers, that*s already working very well for us. And so the security realm is a very hot area and one that we think your breakthroughs and ours need to come together. Another huge area that I*m very excited about is moving towards a more natural interface, moving so that you don*t just have to use the keyboard. I mentioned reading off the screen, I mentioned the Tablet PC with ink. Ink recognition is a problem that -- there were a set of companies about eight or nine years ago -- that came out with products that were called, what, "pen computing" products. And they were interesting, the demos, like all these things, the demos worked very well. But then when you figured out what the battery life was and the recognition rate was and the clunkiness was and the parallax was, there were just dozens of things that meant that that kind of burned out. Well, we*re very patient and so we kept our research on ink recognition going full bore and finally about 15 months ago, came out with this Tablet PC that*s the first product based on that. We*ll have a major update of it this summer and the software gets a lot better and the hardware is also evolving at a very rapid rate. This idea of note taking is really catching on, the idea that it*s a small extra cost to get the pen in there. One of the reasons that handwriting is a bit easier than speech recognition is that your recognizing of text is a conscious activity, so if you see that the way you*re drawing the E looks too much like a C, even if you*re not explicitly doing it, you will loop that thing a little better for it the next time. And if you look at when we make mistakes, you can say, *Yeah, even I would have a hard time recognizing what was written there.* Now, with speech it*s not as easy. Speech is another one that will be solved, and will be solved for a broad range of applications within this decade. We see it today for small vocabularies, but not for dictation, not for really important things. You ought to be able to just talk to your cell phone and navigate the information you care about, and that certainly will become a reality. The things that are holding us back from that is we can compare computer recognition to human recognition when the words are randomly chosen and there*s no noise in the environment and when the microphone is perfect. If you take those three idealistic assumptions, the difference between computers and humans is actually very small. But then as you relax those things and go to real world microphones, lots of noise in the environment and you allow there to be context, that the human has a much deeper understanding of the likelihood of words in a particular discourse. And then in today*s computer systems, you start to see a huge gap. And the gap is big enough that even though people do start using these things, unless they have repetitive stress injury or something that makes the keyboard unattractive, they*re not often long-term users. But we think that*s starting to change. We*re seeing in, particularly in China and Japan with our latest software, which, of course, are markets where the keyboard is not quite as effective because you just have big alphabets, thousands of characters. And so, when you*re using a keyboard there*s a level of indirection between those keystrokes and the alphabet. We had a contest in China where we were able to beat the best typist to get to a perfect set of input by starting with speech, and so that*s the kind of milestone that makes us very optimistic. Vision: these cameras are cheap, a $50 camera, CCD array, has very good resolution. And the idea of seeing what*s going on in the meeting, taking viewpoints, seeing what was up on the blackboard, being able to present that as a time sequence, all that takes is the camera. Understanding viewpoints and social clues, we can*t just take that raw video feed and send it out. That*s not what people are interested in. In fact, if you warp the room, you can actually make it appear that everybody is a co-equal participant instead of the kind of views that video conferencing has typically provided. Now, the ultimate in computer science advances is the field of artificial intelligence. And here again, our respect for the human equivalent, the natural equivalent, grows as this proves to be a very tough problem. The actual products in the marketplace that use AI are things like little vacuum cleaners that try and steer around your rug. So we*re right down on there on the rug, trying to find our way around in terms of applied AI. It won*t stay that way. These Bayesian modeling systems and other approaches, we*re starting to use them in things like games, where when you play with a computer opponent, because we*ve watched across the network all these different playing styles and strategies, we can make the computer as good or as bad as you want it to be, and make it incredibly diverse in the way that it*s interacting with you. And so, the fundamental work at the Bayesian level and the understanding systems, we see a lot of progress there. Now, computer science will start to touch the other sciences in a pretty deep way. The best concrete example of this is some work Jim Gray, who works for us in Microsoft Research, did, collaborating with a set of astronomers, including some here at Cornell. And the idea was to say that astronomy had moved beyond the idea of just staring at a lens late at night and being lucky enough to see a supernova and writing up a paper about that. It*s moved to where you need to take the whole corpus of data of all the observations done over time at various wavelengths and resolutions, and propose theories about densities or distances or dark matter that are consistent with that observed data. And this is not a classic database problem, because the information is very disparate, and so coming up with a schema and ways of navigating, creating these Web services, is very much a state-of-the-art problem that you have to involve domain experts in terms of what classification is very interesting. But they*ve made enough progress on this that it*s clear it*s got momentum, it*s happening, and there will be essentially a logical database that theoreticians in the field can sit there and pull and advance their work. In other sciences, the amount of data is even greater, biology being perhaps the most difficult, but also the one with the greatest payoff. Certainly, that*s a field that, like computer science, will be changing the world in some exciting ways, because advances are coming along. And all of these rich visualization, modeling, data mining techniques are very important. In fact, people with computer science backgrounds I think will be very key to all the advances, because systems-type thinking is very important. And, in fact, with my dialogue with the faculty this afternoon, the emphasis on these multi-disciplinary approaches and the excitement around that was very impressive to me, because I think that*s going to be critical and allow for all the sciences to benefit from these tools. Now, our industry is delivering all this magic. The prices go down, the number of people using it goes up in a lot of ways. But what we*ve got is important enough even in terms of today*s system -- and more so as these systems become more effective -- that the idea of making sure that everyone has access, the so-called digital divide issue is a very important one. Getting the prices down, that*s part of it. Broadband costs are actually the biggest inhibitor today, but, as I said, various peer techniques using mesh software that we and others are working hard on, combined with new modulation techniques, ought to really break the bottleneck there and make that something that*s very accessible as well. Here in the United States, a combination of my foundation and Microsoft, as the president mentioned, have done a project of getting machines out into libraries. And at first when we piloted this, we were a little concerned that kids would come in and maybe not do the most wholesome things using the system, that the systems would break, that the librarians wouldn*t like them, a lot of concerns. In fact, over a six-year period, with the right training, involving lots and lots of people, this thing has been a phenomenal success. In fact, now in every library, there are now 50,000 new computers connected up to the Internet with the latest software. And the librarians are seeing more traffic coming in to actually check out books as well as use the computer, so it*s reinforced the role of the library as a focal point in the community and a place that provides equity, so that the kid without the machine at home -- if he can get to the library -- he*s got that leveling factor. Getting this technology into education is a major challenge and one that I think is very, very important. Getting this technology into poorer countries, there are particular problems, even things like power not being as available in various rural villages. As people think about this field globally, there*s a lot of concern now that not only have transportation systems enabled manufacturing jobs to be done anywhere on the globe, but they*ve enabled all jobs, including jobs that require college education or just jobs answering the phone, they allow them to be done in different places. And this is going to create a lot of opportunity, it*s going to create a lot more effective goods and services. It*s sort of free trade brought to the next level. And I think for the U.S., I*d label it as more of an opportunity than anything else. We need to strive to keep our edge, which is by doing research. Certainly for Microsoft, the lion*s share of the work we do will continue to be here in the United States. We*ll grow outside the United States, but we*re not cost optimizing for doing Windows for five percent less; we*re optimizing or having those breakthroughs come 10 percent faster and the quality be that much better. And delivering it to the most demanding market there is, which is the market here in this country. In the 1980s, it was fascinating. There was all this angst about Japan, and Japan taking over various industries. And some of the humility and thinking that came out of that actually led to the great work that we saw the benefit of in the *90s. So I*m hopeful as we look at the fact that the Internet, software and hardware are enabling global activity, we*ll go back to basics and say, *No, we don*t want to close the door, but we want to make sure that we*re leading the way and that we*ve got our own unique contribution to that picture.* One of the great challenges is in education, to make sure that the quality of education at all levels is super good and to make sure that the entire population is participating in that. If we look at the engineering and the sciences, there the progress is not as strong as in some of the other professional fields. I*ve had a chance, through the Millennium Scholarship Program, to support a lot of people, a lot of minorities in going into fields that without that would have been more difficult for them to do. And I was pretty impressed to see that here at Cornell there are over 50 Gates Millennium Scholars, so that*s a real endorsement, that these people who have the scholarship can literally go to any college in the sense that it*s all financed, and such a high number have chosen to come here. Computer science, I*m saying very explicitly, is the most fun and interesting field. In fact, if you think of other fields, they*re just not going to change like this. They*re not going to take a device that*s blind and can*t talk and can*t do anything really, it*s so limited today, and over the course of just the next ten years, tackle and solve many of these very, very tough problems. And people who understand those things can really be the ones who participate in the advances in the other fields. So it*s fun stuff. The type of jobs that are available are quite broad. I think we need to do more to get the word out about the opportunities and the range of things that go on. And I*m excited to see you all here. I think all of you have a chance to make contributions to the breakthroughs that I talked about, and I look forward to seeing what you*re able to achieve. Thank you. (Applause.)