빌게이츠 speech 중 인상 깊은 것 한개. 그는 dreamer이며 visionary이고, 실천가라는 점에서 배울 것이 많은 사람.
본문은 미래에 대한 그의 통찰과 동시에 실현되어야 할 수 많은 아이디어들을 담고 있다.
강조 on 웹 서비스 표준, XML, 신뢰할 수 있는 보안, R&D(특히 리서치와 그 결과를 개발로 이을 수 있는 능력과 환경), 자연의 학습 알고리즘, 그리고 사람, 사람, 사람.
ps. 참고로 A4용지 18~20장 정도의 장문임
Remarks by Bill Gates, Chairman and Chief Software Architect, Microsoft Corporation
University of Illinois Urbana-Champaign
February 24, 2004
JAMES J. STUKEL (President, University of Illinois): One of the pleasures of being president of a great university is the opportunity to meet extraordinary men and women who come to our campus to inform, to entertain, to provoke, and to instruct our community of scholars and students learning to be scholars. In my guess, Bill Gates will do all of these.
We at the University of Illinois are very proud to be the first stop on Bill Gates' Five Campus Tour. The other schools are Harvard, MIT, Carnegie-Mellon, and Cornell. I think the students there, your counterparts in Cambridge, and Pittsburgh, and Ithaca, are as excited as we are to hear Bill Gates share his perspective on issues and computing, and perhaps give us a glimpse into the future.
Bill Gates speaks to students and faculty at the University of Illinois Urbana-Champaign, Feb. 24, 2004.
Before I introduce him, I would like to remind you that computing is deeply embedded in the culture of this campus, and we are proud to say we have maintained our edge. Our students, faculty and staff enjoy more than 47,000 network connections by which we connect to the world, and people connect to us. More than 1 million times a week, people log on to the online catalogue of the University of Illinois Library, which is the third, only to Harvard and Yale, in size of its collection. And this campus is a giant in research and development in science and engineering. We have more than 80 centers, labs, and institutes where important, life-altering work is underway. Among them is the widely known National Center for Supercomputing Applications, which is helping to build the future of high performance cyber infrastructure. And this new office here at the far edge of the campus is the Beckman Institute for Science and Technology where 600 researchers collaborate, and finally I would be remiss not to mention the investments in R&D brought to the happy place of having two of our faculty members win Nobel Prizes.
As you know, they are Paul Lauterbur, who was awarded the Nobel Prize in Medicine for his ground breaking work on the MRI, and Tony Leggett, Nobel Prize winner for pioneering theoretical work in understanding super fluid.
But that's enough about us, it's time that we move on to our guests this evening. You are here to see Bill Gates, the Chairman and Chief Software Architect of Microsoft Corporation. As you know, Microsoft is the worldwide leader in software, services, and Internet technology for personal and business computing. Last year's revenues topped $32 billion, and the company employed 55,000 people in 85 countries. And Mr. Gates is an iconic figure in contemporary computing.
While attending Harvard, Bill Gates and his childhood friend Paul Allen started Microsoft, and launched a revolution. The fledgling company was more interesting than the classroom for Bill Gates, so he dropped out in his junior year. In his case, it was clearly a great decision. He not only built a company, but more importantly he built a vision. Both were built on the idea that the computer would be a valuable tool on every office desk, in every home, and that software was key. The penetration of personal computing in our businesses, our offices, our public libraries, on the train or on the plane, and in our home is astonishing, and truly reflects the Bill Gates' view that if the software is right, they will come.
Bill Gates also is an author of two books. One of them, Business at the Speed of Thought, is available in 60 nations and 25 languages. It shows how computer technology can solve business problems in fundamentally new ways. By the way, the proceeds of both books are donated to nonprofits that support the use of technology in education and skill development.
Since he is a man on the edge, it makes sense that Bill Gates also has invested in biotechnology, one of the most exciting frontiers in science, and you probably have heard that he and his wife Melinda have endowed a foundation with $24 billion. Their generosity extends to global health, technology for public libraries that serve low income neighborhoods in the U.S. and Canada, and a variety of other community and special projects. He's an avid reader, a golfer, and a bridge player. He is a household name, a visionary, a philanthropist, and tonight he is our guest. So please join me in giving an Illinois welcome to William H. "Bill" Gates.
BILL GATES: Thank you. It's great to be here this evening. They told me I couldn't come too early in the morning or the computer science students wouldn't be up to hear what I had to say.
I want to share some of the exciting things that are going to happen in computer science, and how that's going to change the world in a pretty profound way. Computer science has done a lot over these last 25 years, but I would say that the most exciting years are the years ahead, and there's amazing opportunities for all of you in contributing to that.
It's great to be here in this particular location. The University of Illinois has a great history of contributing to engineering and the sciences, and actually this is the university that Microsoft hires the most computer science graduates from of any university in the entire world.
I'm always a tiny bit embarrassed speaking in university groups, because I, myself, am a dropout, but I'm not here to spread the word about becoming a dropout. In fact, quite the opposite. I'm going to talk a little bit about how computing got to where we are today. The early days of computing were very big machines, and although they were visionaries like Vannevar Bush, who as long ago as 1945 wrote about the Memex machine. Most people thought of them as tools of large organizations, and certainly when I was in high school the computer was a very daunting thing, people talked about taking those punch cards you get in the mail and putting staples in them so you could defeat that evil machine that was always sending you bills that didn't seem to be correct. And nobody thought of it as a tool of empowerment.
It really took an amazing breakthrough in chip technology, the idea of putting at first thousands, and eventually millions, and in the future billions, of transistors on a single chip to get this idea that computers could become a tool for the individual. I think it's fair to say that personal computers have become the most empowering tool we've ever created. They're tools of communication, they're tools of creativity, and they can be shaped by their user. New applications are coming out all the time. Now there's a few key elements that allowed that to happen. From a software point of view, one of the problems in computing was that the machines from every different manufacturer were incompatible. IBM made different machines than Digital Equipment, which were different than NCR or Wang, or UNIVAC, or all the big computer companies of the 1960s and 1970s.
One of the unique things that Microsoft, myself and Paul Allen, had in mind was that we wanted to have a software layer that would hide the hardware differences, and allowed people to invest in software applications, knowing that they could be used across all those machines. In fact, the goal was to create this virtuous cycle that, as more applications became available, people would be these machines, and, as more people bought them, the economies of scale would allow the prices to come down, creating a thriving personal computer and software business. That was our dream, and that was the thing that got me to leave university and start the company. And it's a dream that to some degree came true. Today 600 million people get up every day and have personal computers that they use in a very rich way.
There are still a lot of milestones in that progression. The very first machine, the Altair, was a kid computer that could only light up the lights, it was a miracle if you could even program it to do that much. Then there was a generation of computers like the Commodore 64, the Apple II, the TRS-80, and Microsoft wrote the software which was inside those machines. It was actually a programming language called Basic that let you get in and play around a little bit with the graphics, and write applications.
A major step took place as we moved to larger memory machines in the early '80s, the so-called IBM personal computers with MS-DOS. That machine by today's standards is unbelievably primitive, slow, very limited storage, but it really created the path for this virtuous cycle to take place. It was in the early 1990s that we moved up to graphical machines. This was an approach, of course, that was pioneered at Xerox's Palo Alto Research Center. And then Apple, with both their Lisa and Macintosh got behind it. We got behind it, putting Windows software on top of the PC hardware. It's hard to remember now, but when that was done it was considered a crazy thing. People thought graphics interface was slow, it was hard to program to. And of course, today we take that completely for granted.
The late 1990s were another step change in how we think of these machines, because that's when they all began to be connected together. The standards of the Internet, the pioneering work done here on the browser as a way of visualizing the information across the entire Internet. Those things created a phenomenon that was quite unbelievable, and a phenomenon that created almost a Gold Rush atmosphere. The number of start ups, as we look back on it, was pretty wild. The valuations of companies that had no business model was pretty wild. But, in a sense that hyper-investment, and that attention all accelerated the installations of the connections, and getting people aware that there was something pretty phenomenal going on here.
Today I think we very much take it for granted. Certainly when I want to look up what's new in some area of science, medicine, I want to look up something about history, I just take it for granted that I can go and type in a few simple search terms, and immediately be connected up with the information that comes from the very best experts in the world. So we've come along way.
In fact, the original Microsoft vision of a personal computer in every home, and on every desk, we've gotten a trajectory that's going to get us there. The systems we have today are not the ultimate device. They're not as reliable as we need. They're not as secure as we need. They're not as easy to use as we need. In fact, we have a technology that we call Watson that lets us monitor, if people are willing to send us the reports, when you get error conditions on PCs. Maybe some of you have seen that dialogue that comes up and says, do you want to send this report in, and that gives us a very statistical view of what drivers, what applications, what's going on in terms of that user experience. So it's one source of data that says to us that we have a long way to go to achieve the vision of the personal computer that's as easy as it should be.
At the same time, people are far more ambitious about what they're doing with these machines. We have a whole new area called social computing, the idea of being able to reach out, connect with friends, meet new people, and ways that's taking place. We have new forms of communication, so-called blogging, and Wikis that are drawing people in to participate in new ways. In the area of entertainment this idea that you can play games with your friends, have massive multiplayer games, not just play but also talk to them, in some cases see them, those things are bootstrapped now, and eventually we'll just take those for granted.
One of the things that helps us drive forward is that hardware advance. The chip advance, as predicted by Moore's Law that says you'll have a doubling in power every two years. And that has held true for these last 25 years. And it looks like it will hold true for the next 10 to 15 years. Actually mapping that increase in transistors into computer performance turns out to be a very tough problem. As we get more transistors and very high bandwidth, we're still limited by the actual delay in these systems that's at every level of the hierarchy. It is very much a limiting factor, and there's a lot of clever things we're going to have to do on this. But, certainly we'll have a lot of transistors.
The graphics processing units, the CPUs, all of these things are becoming phenomenally effective. We have 64-bit computing that will give us an address space that will last us quite a long time, moving up from the 32-bit address space.
And when we think of storage, the limitations of the past where you could literally type and fill up a hard disk, that simply can't be done now. In fact, the hard disks that you'll have by the end of this decade, you'll be able to store thousands of movies, tens of thousands of photos, and everything you create in terms of typing your entire lifetime on that single storage device.
That third path is going up even faster than chip performance. Those double every 14 months or so, and they're literally coming to software companies and saying, what are we going to do with all this storage? What kind of applications, what things can you create that would take advantage of that?
Screen technology is another very key factor. We eventually need screens that have unbelievably high resolution. There's no reason that things like magazines and newspapers should be delivered in paper form. The cost, the inability to search, to annotate, the fact that it's not completely up to date, all those things are much superior in digital form. But, our systems still require batteries, and they're still fairly heavy, the resolution is still improving. But where we'll be in a few years at a crossover point where most consumption of media will move to that pure digital form, partly because of these low-cost LCD screens. A 20-inch LCD, which used to be a $2,000 thing, is coming down, will be down to $400 or $500 price point within three to four years.
And so, we have to think about how we take all that display space and resolution and use it on behalf of the user. And so you have to be fairly adaptive because the display space you'll have at your desktop will be much greater than you'll have as you're moving around. The tablet type machine that you carry simply won't have that same display surface, although at some point we may get screens that literally go back to the papyrus where you can unroll them, and then we can get back to having really big screens anywhere that we go.
And then to graphics processors, those are achieving a level of performance that will let us provide high definition realism as part of a serious software activity, or just as part of the communications or game playing. The next generation of video games will be thought of as the high definition devices, including realistic scene that are already pretty good on today's Playstation 2 or Xbox. There's more than an order-of-magnitude improvement that comes in that generation, and is therefore at a level of reality that will draw people in, and allow for game genres that really haven't made sense to this point.
All of these things will be connected with very high performance wireless networks, you're experimenting with this in the Siebel Center, I know, but things like ultra wideband will provide hundreds of megabits of connection. And so the idea that you have to connect the computer up to the display, that will be very antiquated, you will connect up to the display simply over that wireless connection.
And various new approaches like Wi-Max will let us deliver wireless data in a very low-cost way without building a lot of infrastructure. That's fundamentally important to get computing out into all countries, where you can't afford to run fiber optics, or DSL, or cable-modem type infrastructure into all the residences, but these wireless technologies, taking advantage of semiconductor advance in the spectrum will give us essentially infinite spectrum to those homes at very, very low cost. And so that's a breakthrough that we're just taking for granted in designing in to the assumptions we have about the software.
There will be devices of all sizes. The screen that's up on the wall in a meeting room or in the living room in the house, that's your largest way of interacting. You do that at a distance. I mentioned the desktop, I mentioned the Tablet. Of course, the pocket-sized devices are getting far more powerful as well and the idea that your digital wallet, GPS locator and games and personal information will be there, together with your communications functionality, we'll just take that absolutely for granted.
We've even moved to a device size somewhat smaller than that. We've come out with actually a watch that I have on here. This is called the SPOT watch. And what this does is it receives a data signal over the FM network. It's a data-sideband approach. And so as I just look at my watch, not only do I see the time, but I see my calendar that's kept up to date, I see news, I see weather, stock prices. I get instant messages from people that I've authorized to send me information right there on my wrist. Sports games, you can see while they're in progress who's on base, what's going on, and then get the report on anything that you're interested in.
And the chip that's in here, which is an ARM microprocessor, has 10 times the performance and 10 times the memory of the original IBM personal computer. And so we can literally download programs into this device over the FM channel, we take what are called CLR programs and send them to this thing and so we can always create new channels, new ways of gathering information and it's ubiquitous and secure.
And so scaling on to all these devices and getting them to work together, so your information shows up where you want and you don't have to manually synch these things or think about the information mismatches, those are big challenges and those are software challenges.
In fact, software is where the action is. I admit to some bias in this, but I think even objectively the hardware people are doing their jobs, they are going to give us the miracle opportunities, but will it be usable, will it be secure, will it be fun and exciting and approachable? That is purely something that the software industry needs to deliver on.
Let's look at different domains where software can help us be more effective. First, let's take people at work. People at work overwhelmingly are what we call information workers, designing new products, dealing with customer service, forecasting demand, buying and selling. Those are the kinds of jobs that overwhelmingly in developed economies are the vast majority of people.
And competition exists in terms of how effectively you do those jobs. Do you design a new model properly? Do you anticipate the demand? Do you understand the least cost way of getting something done? Do you see where your quality problems are? And the insights into those things can be provided through software.
The lack of visibility of what's going on and all the information about a business that people have today is really quite unbelievable and they don't have the expectation that they should be able to look at all those transactions and data mine the transactions and navigate the latest information.
But software can change that. Visualization techniques, modeling techniques, even things that you might think of mundane, saying that, hey, when you have a meeting let's make that meeting 20 percent more efficient, let's allow people who aren't physically present to participate in a very rich way. When you have a phone call, why can't you just connect your screen up to their screen so instead of talking about a budget or a plan or whatever the information is, you can sit there and edit that together?
The very mechanism of capitalism, finding buyers and sellers, there was a lot of hype in the late 1990s about how that would change and become friction free, but, in fact, the software infrastructure was not present. The idea of having software anywhere on the planet being able to find other relevant software and exchange very complex information, we didn't have the protocols, standards and tools to make that work.
So as we connected up things to the Internet, we connected them up with a presentation standard, HTML, but the idea of arbitrary software, no matter what the application is, but take buying and selling as a good example of it, we don't have that today. And with the challenges of security and things, that's not an easy thing but it is being built.
These are called the Web services standards, and they're fundamental to letting information be exchanged in a rich way. They fulfill a dream of computer science that existed for a long time, dreams about heterogeneous information, that the advances in XML are finally solving those very tough problems.
And so within the next year, as that foundation gets into place, a lot of those dreams of the late 1990s will become a reality. The cost of a transaction, the cost of finding who can well you the product that is absolutely the most suitable and check their reputation and check the state of that transaction, all of those things will move to be digital, and that hasn't happened yet but with the software advance that will absolutely take place.
People waste a lot of time on various communications modalities. Today software doesn't know which calls or e-mails are important to you. We've all been in meetings where peoples' cell phones ring. We've all gone to our e-mail and found lots of unusual, unwanted e-mail that wastes our time. I have been offered many university degrees in that spam e-mail. (Laughter.) I don't know if they're targeting me or if other people are being offered those as well. The most interesting ones, they said that for dollars a month they would pay all my legal bills. (Laughter, applause.) That one, I know they didn't mean it to come to me probably. (Laughter.)
Another good story about that is just this weekend my wife and I were sleeping in a little bit. Our 7-year old came in and woke us up and said, "You've got to come, you've got to come." And we said, "No, no, no, it's still 7 o'clock, why don't you go back and keep doing what you were doing?" And she said, "Well, I was using the computer and it's amazing." And I said, "Well, keep using it." (Laughter.) And she said, "No, no, no, we won, we won money, dad." (Laughter.) And I didn't want to say something flip, like, "Hey, we don't need more money." (Laughter, applause.) So I got up and, of course, it was one of those come-on type things, and there's my 7-year old who thinks she's won some amazing contest, and I'm trying to explain to her about it's just somebody trying to get her to go to that website and all that.
So we have a lot of work to have the computer model our interests, what is worth interrupting us for at various contexts we're in during the day, what kind of e-mails should we see no matter what's going on, what should only be brought to our attention as we go home.
How do we organize our tasks? Think about all the different things you want to get done; the computer is not very good at helping to organize those things, notifying us about deadlines.
Literally take phone calls today. If you call somebody and they're not available, if you can prove who you are through some caller-ID-type mechanism, if you're a person who works with that other person, their software ought to negotiate with you, looking at your schedule, to find exactly the best time for you to meet or be in touch with each other, and the idea of phone tag or busy signals and those things should really become a thing of the past. But we need a software model. We need something that's adaptive, that learns, that has the right authentication built underneath.
And we have far too many communications things: e-mail, phone and, even phones, we have our phone at home, we have the portable phone. The fact that we have to remember phone numbers and update those things, the instant messaging is a world of its own; all of those things really have to come together and help people and make people far more productive.
In terms of things that people do at home, we are at the beginning of a revolution in terms of people being in control, control of when they want to watch a TV show that the digital video recorder is now getting people addicted to this idea that it's up to them to decide when they want to do it. People are getting addicted to the idea that, in terms of their music, they can organize their collection and have different play lists, that they can have a portable device that they take with them that lets them play that music.
We're even getting to the point now where we can take videos and put those on a portable device.
This is a little device called the Portable Media Center. You can see the basic size of it and that shows what comes up on the screen. You connect this to your PC over a wireless or a USB cable and you can take whatever TV shows you recorded, your movies, your pictures and all of those things can be downloaded onto this hard disk. It's a 40-gig hard disk, which, of course, is becoming unbelievably inexpensive, and then relative to a music player the only extra expense is just having this LCD screen, where that too is becoming quite inexpensive.
And so this is a different way of thinking about consuming media, putting the person in control, having it wherever you want it, having your lifetime collection easy for you to get at and work with.
And as people have all this different media, we need to make it easy for them to navigate around in this information.
I've just got two little quick demos that are ideas coming out of Microsoft Research that give a sense of how we think visualization can be made a lot better than it is today. The first screen I've got here is to help you look at a set of movies or a movie collection. And so at the center we have a particular movie, "Blade Runner," and you can see that off on the side here it takes things that are related in some way, like everything that's directed by Ridley Scott, it shows and I can go in and cycle through at any speed, see those different things, and I can pick one of those and say, OK, put that at the center and then go look up in the database, get me the information and tell me who are the actors. So here are all the Anthony Hopkins movies, here are all the Julianne Moore movies. I can pivot there. And so this idea of going back and forth between these different things becomes a fairly straightforward thing.
Another example is dealing with lots of photos. This is a case where it's going to be so easy to take photos, you're going to have thousands and thousands. And, in fact, one of the researchers at Microsoft Research goes around with what she calls a little photo button, and it's noticing transitions during the day and it's taking a few hundred photos. And so she doesn't even have to think about actually clicking a camera; she just gets at the end of the day all these interesting photos that she can decide if she wants to share with people or in terms of having memories about her activities or things she's doing with kids or friends or things like that, it's there at no effort at all.
Well, you're going to get a lot of these photos and what do you do with them? Well, this is a research project called Media Frame to start to suggest that we can have user interfaces that make this practical.
So you see we have a bunch of images here, hundreds, we can hover over different ones of these. And some of these actually aren't photos, they're actually movies. It's our belief that you'll more and more not think of photos by themselves and movies by themselves, but rather you'll think of still images, motion video and all of the audio that you capture either at that time or that you can easily add later on, we'll think about these things as wanting to organize them together.
Now, sometimes what you want to do is put various keywords on these things and you can see here we've done that a little bit. So let's take one, let's go in and look at the thing that relates to Thanksgiving.
I still have a fair number of photos here, so I can go in and use a software algorithm that shows me which are the ones that have faces in it and those get highlighted, or which are the ones that are indoors and you can see it's automatically able to tell which those are and highlight those.
And so we have recognition software that actually did the orientation. It found and notified me of all the slides that were coming in mis-rotated; it did that without my having to spend time scanning through those things and it can see these different photos.
And, in fact, if I take the photos of those faces and I tell it who somebody is, if I make an association with my contact list, then in the future it will be able to do that recognition and do that categorization in a very automatic way.
We have the idea of finding similar images. Actually, let me go back into that and try the similarities. If images are similar, it's actually looking at what's inside here. And so if I can take this image and say, okay, what else is similar to that, if I relax the constraint, eventually everything is similar, but at this rating it's just these particular images. And so actually intelligent analysis is part of how we'll be able to deal with these things.
If we go back and see the whole set again, we can also try out a different view where we're using 3D. And here what it does is it takes and organizes them by time. Of course, the camera is storing lots of metadata with these photos. It has a clock in it. It is able to know when that photo is taken and I can just switch and change that X-axis and break it down into different groups. And as I select a group of photos, I can use these tags, add tags, change tags on a whole set, all at once.
And so this is just an idea that we ought to be able to make it reasonable to play around with lots of different photos and media clips and make navigating through those things a very, very simple activity.
Well, the wellspring that really drives software forward is research and research is done both at universities and in commercial organizations. And, in fact, the United States is dramatically the leader in both aspects of this. The best universities are overwhelmingly here in the United States doing this work and there's a real symbiosis of the relationship between the companies trying to build these things into products, whether they're startups or larger companies and the universities; very much a virtuous cycle of sharing ideas, helping research get funded, creating jobs for people and it's worked in a really fantastic way.
Microsoft is a big believer in investing in R&D. Our R&D budget at $6.8 billion is substantially the largest of any technology company. And it's kind of amazing to me, when I grew up, I always thought IBM was the big company and actually in terms of employees they are the biggest. They still have 330,000 -- I shouldn't say still -- employees, because they've taken an approach that's more based on services and doing different things than we do. We're very focused on building software products, but to do that it's got to be about R&D, and R&D that looks well out into the future and takes on the very toughest problems.
There are some good examples of collaborations here at the University of Illinois. The Gaia.Net distributed OS was something that some of our devices and software components can come in there, and I'm sure we'll learn a lot from what's going on there.
The experimentation of the Siebel Center, built on a lot of different kinds of software, including some of the Conference XP things we've done there, we're very excited to see what can come out of that.
Some of these research problems are very tough problems. A good example of that is what we call Trustworthy Computing. In fact, when I met with the faculty earlier, I was very pleased to hear this is going to be a major focus of bringing together a lot of research ideas about security and reliability into an institute that looks at it in a very broad way.
When the Internet was first designed, it was designed assuming that different parts of the network could be malfunctioning, that they might be broken or literally that they might be bombed, but there was not an assumption that there would be malicious actors on the network, and so there's no authentication of the From and To addresses. SMTP mail, there's no authentication of who that mail is coming from. Many of the software systems are built around passwords that are truly a weak link in terms of being written down or used on less secure systems or being very guessable.
And so what we've ended up with is a situation that's very fragile. Any software bug can result in what's called an escalation of privilege, and then hijacking a system to either flood the network with traffic or to send lots of e-mail out that appears to come from that person or various kinds of attack methods that are taking place.
There is no doubt that for computer science to fulfill its role in helping business, helping entertainment, that we've got to make this network secure and reliable, that we have to be able to make privacy guarantees to people in terms of how information is dealt with on this network.
And there's a lot of invention taking place here. This has been our biggest area of R&D investment for many years now. It was about three years ago that we really pushed this up to the top of the list and really brought in a lot of additional expertise.
Some of the issues are very simple to solve: Moving to smart cards instead of the password, having software be kept up to date so that when there are problems they don't sit there so people can do exploits, having firewalls so you partition the systems off and you don't just look at what type of remote call is being made but you also look at who's making it, the transition to IPSec and IPv6 will help us with this.
There are new programming methodologies around interpretive systems, like our Compact Language Runtime, the CLR, that helps you define the privileges of a piece of software so you're not just doing exactly what that user is privileged to do but rather saying what's appropriate for that software.
Some of the newer techniques have biological inspirations of monitoring systems and having a way of looking at them and seeing when their behavior becomes abnormal, both at a system level and at a network level. So that's a very exciting area.
Another big area of investment is what you might broadly call natural interface. The keyboard is okay, the keyboard is going to be around for a long time, but it would be far more natural if we could use ink and speech as a way of getting information into these systems.
These are tough problems. They've been worked on for a long time. Ink is somewhat easier than speech, partly because users have a very explicit model of what readable handwriting is and what it's not. And so even as people start to use our Tablet PC that's got ink built-in, they find themselves taking an E versus a C, being a little more careful after they've had recognition errors, to loop the E and not loop the C and so you get more and more accuracy.
And so these handwriting systems are really coming into the mainstream. The cost of the digitizer is extremely low and the way that software is adapting to it, we'll take this for granted that every portable PC is a Tablet-type PC within the next two to four years.
Speech has been a little tougher. It's one that we are investing in very, very heavily, but users have no explicit model of speech. In fact, when the speech systems start to make errors, their tendency is to not only get irritated but talk louder. And whatever the model is of their speech becomes less and less capable as they're getting slightly more irritated at the system. And the fact that there's no predictability and the system makes errors that every other thing you've ever spoken to, which are humans, would never make those errors, is kind of frustrating.
And so we have to get the accuracy levels to be extremely high. There are great advances here, not just driven by the extra power we have, but modeling, going through, for example, all of the user's e-mail and understanding the corpus of words that are typical in their discourse, we're using that both in mail and in speech capability, having deeper language models, having better microphone type systems.
One thing that's fascinating is that the difference between human and computers, in a noise-free environment -- well, if you take the best case, a noise-free, context-free environment where you're just doing random words for a human and a computer, the computer is not that bad, the difference is very modest. Where the human gets the wild advantage is that the human has context, they have a sense of what the speaker might say next, based on what's going on and what they know about the subject. And humans are dramatically better at doing noise elimination and this is a case where the signal people and the speech people are coming together now to get a sense of, okay, how does the human audio system do this.
Like most things related to human capabilities, our appreciation for how good the human system is just gets higher and higher as we try and create the equivalent on a digital basis.
The ultimate natural capability is the idea of artificial intelligence, and there is less research on this today than when I left school 25 years ago, but there is some very good research going on. Bayesian systems are a type of system that attempt to model non-linear activities, and there are many similar approaches that are becoming ripe and can be applied in interesting ways.
We're starting out with some very simple things. The only AI product that actually sells today is this vacuum cleaner that goes around, so that gives you a sense that we're really at the low level there, down on the rug trying to find our way around.
The next generation will be using AI algorithms in games. If you play a computer opponent today, after you've done that for two or three days, that computer opponent becomes somewhat predictable and the range of skills is either too high or too low. And with an AI machine built in there, we'll be able to make that richer and richer, in fact, learn from users how they play, gather that information centrally and reprogram the AI machines down on those different systems.
One fascinating trend is that all of the sciences are becoming very data driven. Take a science like astronomy. Jim Gray, who's one of our researchers, realized that if you want to propose a theory about astronomy, you need to look into all the different databases that are out there, and yet these databases were not connected in a way that you could perform these very rich queries and try and see what's the density of a start system like this or are there any cases of something where these two things are near to each other.
And so he led a project taking very advanced software technology, Web services, and built, together with a lot of collaborators, what's called the National Virtual Observatory. And so no longer is astronomy just sort of being at 3 in the morning with your eyes to the lens when a supernova explodes, but rather it's doing sophisticated data mining and looking and forming theorems about the information that's been recorded over all time in this very large virtual database that's been created there.
That same sort of thing is necessary across all the different sciences, biology being probably one of the most interesting and challenging. And so the interplay between people who have computer science backgrounds, and data mining and modeling and networking and what they'll be able to bring to advancing biology in these next several decades, will be quite phenomenal.
I think it's really biology and computer science that are changing the world in a dramatic way. Other fields, they're great, but they are not changing the world. They're not empowering people. They're not making the advances like these are. And it's actually at the intersection of these two fields where perhaps some of the most interesting activity is taking place.
Certainly nature has come up with a learning algorithm that we have no understanding of, and as we, through various techniques, are approaching that kind of capability, implementing that in software will be a profound contribution.
While all this computer activity is so neat as a tool, one of the big problems we get is the so-called digital divide, that is that you have a lot of people who have access in richer countries but even there, not everyone, and yet the tool, you'd like it to be everywhere. What's the solution to that? Well, you can drive the cost down on the hardware side, the software side; that's happening very effectively. You can make sure there's lots of philanthropy and donations around it; some good activity there. It's actually the communications costs, the broadband connections that are the most expensive part of this, but even there the advances in wireless can solve the problem.
One of the projects I have the most fun with that Microsoft and my foundation did over the last six years is go out to 18,000 different libraries and put in a total of 50,000 computers that are just sitting there so that anyone who can reach a library can get in, get out on the Internet, get that information and use the latest software. And it's amazing to see how people come in and use that.
There's a lot more to be done there in terms of the schools in terms of looking at that on a global basis, but it's a very important goal, particularly if you see this as almost like literacy, like reading literacy in the same imperative for everyone to have access.
Now, the tools of technology are changing global competition and there is a lot of concern about this. The tools of technology are making it possible for not only manufacturing type jobs to be done anywhere on the globe but actual services type jobs, not just programming, not just call center but design, architecture, any type of work, if you have these rich collaborative interfaces that the Internet and the rich software on top of it make possible, that will let people compete for that work anywhere around the world.
And so we're going to go from a world where the thing that would predict your opportunity best historically was, were you lucky enough to be in one of a very few countries, to, in the future, the best predictor will be what's your level of education. If you have a college education, no matter what country you're in, there will be substantial opportunity because of the way these things connect together.
Now, this is an interesting challenge for the United States. The United States actually did its best work during the 1970s and 1980s, and that was actually a period of great humility, of concern about international trends. In fact, the great concern of that era was that Japan had a better model, Japan was ahead of us, Japan was going to own industry after industry, and just wipe out the United States -- including computing was going to move there. And although that was completely overblown, completely wrong, it underestimated the vitality of both the commercial and research side in this country, it allowed us to really step back and examine what our strengths were in driving forward, and that's why such amazing work I think was done during that period.
Here we're going to have that same type of questioning as we're seeing more global trade and all these different activities, as we see particularly India and China stepping onto the world stage with their university output and the energy and the innovation in those countries, a lot taking place, that will challenge the U.S. to say, are we really able to keep our edge, are we really able to keep ahead? And it's the investment in research, the value of intellectual property, it's a lot of things that the U.S. is actually pretty good at, that we just have to renew our commitment to.
So my view is that in the next 10 to 15 years computer science really will be magical, that the impact, whether you think what it's going to do for medicine, what it's going to do for education, what it's going to do for worker productivity, the impact is really hard to exaggerate.
And I'm not saying this is going to happen in the next year or two. Every year there will be some neat things as speech and ink and all these things come along. But it's really the accretion of those things where people are used to the tool and the tool is super secure that creates this shift in how things are done. How will education be done, how will that change? Well, that's one of those great open questions.
The key element in doing this is having great people, and Microsoft succeeds by having great people, universities succeed by having great people and making sure that the re-investment in those people takes place.
There's a little bit of concern that the peak enrollments in computer science are off from the years past, and looking at that, particularly on a national basis, it says, OK, what aren't we doing to show the opportunities that are here?
Another challenge, of course, is the lack of diversity; both women and minorities in computer science are not nearly at the levels that we'd like. Obviously we'd like those numbers to be 50 percent, purely diverse, and yet the numbers are much more at the 10, 15 percent level, and a lot that needs to be done about that.
I'm sure that this is a very multifaceted thing in terms of showing the opportunity, giving people an opportunity at a young age to see that it's very interesting and pointing out that these jobs aren't all just hard-core coding type jobs. There are plenty of those, those are neat, I like that, but a lot of them are much more in terms of having skill sets where you need to know computer science but also understanding usability and social factors and marketing and business and bringing those things together. Those are a lot of the really great jobs that are there.
On the minority opportunity front, I'm very pleased that I've been able to sponsor what's called the Millennium Scholarship Program. (Applause.) Here at this university there are 25 Millennium Scholars, including some here tonight. It's a neat thing and I hope all of you will serve as role models and really encourage other people to do the things you're doing, because I think that's a key part of the path forward.
So the tough problems just take great people. We will have any type of simple user interface, secure type systems, and the direction this is going to head in, there's a lot of unknowns that are going to make this, in my view, by far the most interesting place to be involved.
And so I'm excited that many of you will go through a computer science program and join a variety of companies, perhaps Microsoft, perhaps some of the startups, and really make this a reality, because this is the important stuff and the great stuff is all ahead of us.