MR. GATES: (Applause.) Good morning. It's fun to see every year how elaborate Comdex has become. It's fantastic how people try to draw attention to their booths and do the most amazing demonstrations. We've come a long way since 1985, when I first had a chance to do a keynote here. At that time, I had my dad come down and change the slides for me. (Laughter.) And the only thing I worried about was the condensation on the slides, whether they would dry out properly or not.

It was a lot of fun to hear yesterday from Andy Grove about the 25th anniversary of the microprocessor. It got me thinking back to the early days when Paul Allen and I first saw an article in Electronics Magazine. It was just after the 4004 had come out. I remember the year pretty well because I was trying to get my driver's license. (Laughter.) And Paul looked at this article and said that even though this first chip had so many limitations, he saw that through Moore's Law it would become more powerful than any existing computer. And he and I got very, very excited about that.

Essentially, the extrapolation we did when we saw that article led to the creation of Microsoft, and we decided that we'd focus on what we did well, which was software. Now, the beginnings were fairly slow, the 4004 couldn't do much. Even the 008, which came in 1973, Paul asked if we could do a Basic interpreter for it, but it wasn't possible, so we did dedicated applications. It was the 8080 that was the next generation that was the big step that led to really the first personal computers.

What I want to focus on today is the future of the personal computer. I'm going to do that in a fairly simple way. I don't have any demonstrations, just a few videos, and there's no one else who's going to come out on the stage. That simplifies things for me. I'm never quite sure when somebody comes on stage, you know, how you're supposed to greet them. I mean, the honest thing to say would be, "I'm sure glad you showed up, otherwise I don't know what I'd do next." (Laughter.) And when you're sitting there looking at the demo -- of course, hopefully you've rehearsed the demo many times -- are you supposed to look surprised like you didn't really know what they were going to show you? (Laughter.) So we've avoided all of that today.

I considered doing a new video which showed a vision for the PC, but as I went back and reviewed the different videos that had already been done, I decided that people really had captured very well what some of the opportunities were. And so what I did is I took and created a montage of some of those visions, and let's go ahead and look at what we've all been predicting for the PC.

(Video presentation)

MR. GATES: (Applause.) It's amazing when you look at those videos how many crises there are. Medical disasters. Every once in a while they mix in a classroom scene or a little bit of a business deal. But the world is an exciting place.

In those videos, we see a lot of common elements. The basic view is very, very similar. We see fantastic bandwidth. Almost everyone had some video conferencing where you were in touch with somebody far, far away. Most of them had large screens so that the ability to see lots of information and context was even better on the screen than it is today on paper. We had small machines that were paging people, letting people transfer, digital cache. We had some voice recognition so that you didn't have to go to the machine and do something. And perhaps most importantly, we had agent technology, the idea of the machine adapting to what you're interested in and doing the right thing for you. So all of this is very exciting, and all of it will be possible within the next decade.

I want to split the way I talk about this future into three sections. First, I'm going to talk about the success of the PC. I think there's a lot we take for granted about the PC and how it's changing quite rapidly. The second area is to talk about the PC and the Internet, the biggest thing that's happened to this industry for many years, and really enabling some of these rich things that have always been part of the vision. And finally I'm going to talk about technologies that are very rich, very ambitious, and will make the PC even more personal than it is today. You'll get a sense that I see the PC becoming more powerful, not less powerful.

Now, over these last decades, there have been several times when people have looked at the PC and said, "Hey, there's something missing here, and maybe we ought to throw out what we've done, throw out the applications, the install base of hardware, the industry structure that we've put together and start over with something new." A good example of this is the architectural challenge that came from Unix-type workstations. Remember, Unix workstations for a long, long time were far more powerful. They had things that we could only dream of and were going to take us a long time to achieve: 32-bit, multitasking, rich graphics, high-end applications like engineering applications.

It's only in this last year that we can say that we match those workstation systems toe to toe in terms of performance, application availability, and the richness that we have in our system. And so it's an interesting question, why didn't we, over a decade ago, switch over to use that Unix base? Well, there's a reason that has to do with the basic approach that was taken there, which doesn't tie in to the kind of evangelism and breadth that have been the strength of the PC.

When graphical interface came along, again, the Star, the Lisa, these pioneering graphical interface systems were not compatible with this PC. They were a whole new effort. And yet both of those found limited success, despite the wonderful characteristic of having the graphical interface. And it was sometime, say, 1990 to 1992, that that became a mainstream feature of the PC and now is part of every application that you use.

We've also had various attempts to recentralize computing, to go back to the time-sharing days. Again, this really ties into people who want to sell expensive servers. If you can take the cycles off of the desktop and move them back to the center, then you can have very big and expensive machines that are doing that work. And X Terminal said, you know, we can have a graphical presentation, all you have to do is share in the center. Well, the latency, the richness of that approach meant that it was used in very limited cases.

Another big challenge has been object orientation. I can't tell you how many Comdexes I've come to and sat on panels where people debated who was more object oriented than the other person. In fact, the hype got so strong that they had to invent a new term, which was "true object oriented" as opposed to just object oriented. And there were many proposals to build "from scratch" systems. A great example is Taligent and this was a company that combined the learning that Metaphor (sp) had done with some very smart people from Xerox. It took the strength of IBM, the strength of Apple. H-P came in as a major investor. And it was a new system. But that effort was basically abandoned when they realized that the PC would be able to adopt object oriented approaches where component software could easily be done on the basis that existed there.

Well, as the Internet came along, the same sort of questions were raised. Do we need to start over? Should we take a browser as the starting point and essentially extend that into an operating system and ask that new applications be built around it? And there's been a lot of talk here about, you know, what kind of new device might this lead to.

Certainly the Internet is going to have a huge range of devices connected to it. Telephones will be directly or indirectly connected to the Internet. There's two very important form factors that I think will be popular and yet require you to subset the PC. One of those is the hand-held device where, because of the screen size, because of the cost requirements, because of the battery life, you want to scale down both the operating systems and the applications. And I think there's great progress being made there. We just introduced, on Sunday, our hand-held PC approach using Windows CE, and there's many vendors building those subset machines.

I also think that connecting up through the TV set is quite valuable. There are a lot of people doing this in a couple ways. Using an intelligent set-top box is one approach, using a video game and extending it by plugging in a modem that includes the software for browsing is another way, or you can just have a dedicated box that you connect up to your TV set that has that browsing capability. Microsoft is working with a number of companies on this. In fact, Web TV is one company we've made an investment in, and we're taking our browser technology, a subset of Internet Explorer, the appropriate subset, and putting that in to the TV device.

So all of that is very complementary to the PC. The PC is still the full screen device that you want to sit next to, edit your e-mail, do you taxes, do your homework. It is the full-scale approach.

Now, some people are saying, even at that full-scale level, they'd like to start over and, of course, the term "network computer" is used, but it's a tricky term because there are many different definitions that people have for this. Some people talk about it as the device without a disk. And, of course, by eliminating the disk, you can save a couple hundred dollars. There have been diskless PCs over the years, and in some special circumstances they worked very well. For example, Sony is a partner of ours who's deploying in an airline entertainment application seat-back Windows 95, and those machines connect up through an optic fiber to a server that provides all the disk capability. And because they have a very high-speed connection and because the demands for those applications are not that large, it works very, very well. And so we will have more people taking advantage of diskless PCs.

Now, beyond the disk, it's tough to eliminate a lot because, if you're going to run a state-of-the-art browser, the memory size or the processor capability, the screen, the graphics, the sound, the pointing device -- basically all the things that are in a PC -- are absolutely necessary there. And one of the things to be very careful about here is -- particularly when people talk about diskless approaches -- is that we don't want to overload the network. You know, somebody at home at 28.8, you're not going do paging across that network. Even on an Ethernet in a business, the network will be greatly overloaded if you go diskless, and so the response time for the user will be greatly reduced. Also, you'll end up spending dramatically more than the cost of that disk, in terms of enhancing the server, trying to get the performance up in that kind of time-sharing type approach.

And so, when people say total cost of ownership, it's important to have that server aspect and the extra costs you're creating there, and bring those into the total picture. A lot of what this attack has become is sort of a questioning of empowerment. You know, somebody likes to go around and say that the only key they've ever used, the only command key in a productivity application, is the backspace key. Well, our view of the market is quite different than that. We'd like to give more and more power out with these machines but hide the complexity.

And there is an interesting opportunity here in terms of rich administration; the idea of taking the state of a PC or any device and mirroring that up on the server allows a lot of neat new things to happen. It makes it easy to move from machine to machine and get your custom desk top and profile. It makes it easy if you switch machines to have that state be replicated down. And so that's a very important idea, that we call Zero-admin Windows, that'll be a standard feature -- without changing the operating system, without changing the hardware, without changing the applications, you'll be able to get that.

But we are staying with the PC -- the ability to print, the ability for people to go out and buy a piece of software if they're interested in that. And so there are so many strengths to the PC phenomena that we often take for granted. I wanted to quickly go through a couple of those. Certainly the user base we have is pretty unbelievable. I don't think anyone in our industry five years ago would have predicted that we'd get to this scale. You know, 200 million is unbelievable, and the device sales every year we've seen very, very healthy growth. The growth percentage, of course, is going to come down. When you get to over 70 million units a year, you know, it's hard to grow at greater than 10 percent. And there probably will be some years where there won't be any growth at all, but the basic trend line I think is very strong.

We have lots of hardware developers who participate in this, and that flexibility is a benefit for users. I think there's a certain irony that the Unix platform, which was so closely associated with the term "open," in fact today if you do a side-by-side comparison of cost, of hardware choices, of software choices, there's no doubt you pay an incredible premium and give up a lot of flexibility when you move to what was termed "the open system." The PC approach has proven to be superior, and as PC performance moves up to not only match but exceed what was done there, that industry structure is an incredible benefit.

And millions of software developers are a big part of this, and it's not just the commercial companies that we see here at Comdex. Even more important, I would say, are the corporate developers and the wealth of applications they built with incredible tools. It's always amazing to me every time I get out to a customer and I see how they've used the richness to customize to their particular needs, the way they want information to flow inside their business.

And so, as people are making choices here, there's a lot of flexibility. Just take a single issue like programming languages; you can use any programming language you want. You can use COBOL if you want to, you can keep using FORTRAN if you want to. You can hack around in machine language, or use C, or use Java, or use Visual Basic. All of those things are supported within this architecture.

The economy of scale that we get here allows very advanced technology to move quickly into the market place with very, very little markup. If we just look at what happened over the last year, this was probably the best year ever for PC improvement. Certainly on the processor side, Intel led the way with high-speed Pentiums -- 200 megahertz -- Pentium Pro has started to move into the market place and in fact the demand greatly exceeds the supply, which is a very good sign. And that is a dramatic speed up, even above the Pentium, which was a big advance. The drop in the price of memory that's taken place over the last year is a wonderful thing. The memory vendors may not view it that way, but certainly for users and people wanting to use rich applications, that's made a big difference.

Also the hard disk providers are to be congratulated for their incredible improvement. They actually move at greater than Moore's Law type rates to up the capacity. Today it's hard to buy a PC with less than a gigabyte of storage, and a gigabyte of storage is a lot. You could type your whole life and you'd never fill up that disk. And so it's letting us deal with new things, like the multimedia elements that are becoming more and more commonplace.

So, the strength of the PC, what we take for granted about it, there's a lot there. Now, the PC has been thought of as a low-end phenomena, and of course it started out with the 8086, and it just slowly but surely moved up that performance spectrum. But I think 1997 is a key year; it's the year that we'll make progress, in terms of scalability, and we'll get that message out to the market place in a very strong fashion. PC scalability has been going up because of many, many factors. The extra processor speed is part of it, software improvements are part of it.

Another element is symmetric multiprocessing. Today, PCs with up to eight processors in the system deliver almost linear performance. So as you go from one to four, four to eight, you get almost four times the performance and almost eight times the performance. Work we're doing over the next year should take that up even further, perhaps to 16 processors. We're also doing more to cluster systems together, including an approach called No Shared Clustering, where you simply use the network to coordinate what's going on.

One concrete benchmark of this scalability improvement is that on a standard database bench mark two years ago we were doing 600 transactions a minute; today we're doing 6000 transactions per minute. And within two years, with the things I'm showing here, there's no doubt we'll be doing 60,000 transactions per minute. And so that's a huge improvement that really gets into being able to handle more than even the most demanding corporate transaction systems that are being used today.

During 1997, we'll show systems with many terabytes of storage. Today in the world there are about 15 companies who, with some combination of hardware and software, have databases over a terabyte. Well, of course, through the Internet, when you're trying to track all the clicks and what people are doing, there will be many, many more of these large databases. So the fact that we'll be able to use PC technology to address this is a wonderful step. A billion transactions a day is far more than the phone companies, the credit card companies or the airline companies are dealing with, but Web sites will push that limit and take you even further.

So we're doing more and more with the PC all the time. In fact, if the industry ever decided let's just freeze what we're doing and let's focus on cost reducing, let's say, the, you know, the 1989 PC, and let's focus on taking the things people do with that PC and making those easier, you know, today the price of the device would be under $500, and all those ease-of-use things would be done.

In fact, if you think back to that stage, what were we worried about? We were worried about how hard it was to connect up to the network, all these different protocol stacks that really didn't work that well, didn't work well together. Well, that's basically solved. You know, machines, you plug them into the network, because of extensions with TCP/IP, it automatically gets its IP address, it reports itself up to the server. So network connections, that's something people don't see much anymore. WYSIWYG, thinking about which fonts are bitmapped and what size font used, that was solved by scalable fonts. Printing is far, far better today than ever before.

We have a lot of things that we bring into the computer that have the computer taking over work that we would have done in the past. A term that Microsoft uses for this is IntelliSense. I love the idea that my common spelling mistakes are automatically corrected, or I see where I've made a mistake. And I don't have to run that batch command. It's pretty cool, because if somebody sends mail to me, of course, it notices their spelling mistakes as well. (Laughter.) And with Office 97 we've taken it to the next level, and we have grammar checking, so it uses a little green underline to tell you if a sentence is a little long, or if you're using "their" instead of "there," so those embarrassing mistakes get caught in a very automatic way.

So the progress is there, but people are doing more and more with the PC. And sometimes the complexity, the underlying richness that really has to part of the system shows through.

We went out and looked at some of the promises that have been made about ease of use. We took some of the more interesting phone calls we get into our product support facility, and some of the frustration people have with all this terminology, and made a video that has customers reacting and showing that they really do expect a lot more here.

Let's take a look at that.

(Videotape presentation.)

(Applause.)

MR. GATES: (Applause.) How many people here have ever been frustrated with their PC? (Laughter, applause.) So we have work to do. And one of the ways we've tackled this is by being in a feedback loop; that is, by constantly getting input from PC users about what they'd like to see made better. Microsoft, for example, takes millions of phone calls a year, and every one of those is logged into a database. Whenever we're doing a new product, we go through and say, you know, "Could we design it in such a way that we get a lot less calls?"

And in fact, with Windows 95 and the latest release of Office, the calls per unit are down substantially. But still, we have a long ways to go on this. It's very important that we keep those calls coming. It's a fundamental part of the improvement process. In fact, by handing this Internet connection to customers, we'll be able to get even more of that feedback. Literally, when you're using a feature, if that feature isn't working well for you, you'll just go to the Help Menu and click and say, "I'd like this to be better," and that whole context of how you were using it is sent up to us, along with any comments that you want to make. We're tightening the feedback loop to make that better.

I would say that during the last year, there's been some very good progress, including some things that we talked about doing for a long time that were obvious that needed to be done. You know, what a breakthrough -- long file names. You know, six characters, that was almost enough, but people wanted more, so we did it.

Multitasking, you know, I remember going on OS/2 road shows and talking about how multitasking would change people's lives. Well, it needs applications that use the threading. It needed people to move up to the 32-bit level. And we have been able to achieve that, and so now people just take it for granted. It's there; it makes the PC a better device.

Plug and Play was a broad industry initiative, and it really has made a difference. The auto-configuration capability there is working for lots and lots of users. Now, we need to get down to the details and make sure that for all the new devices this works well and make sure that absolutely everything has been checked off. And that's where the feedback loop comes in. Plug and Play is the right architecture, but the details are still being refined there to deliver 100 percent of that promise.

Auto-play CDs, another very obvious idea. If you take a CD, put it in the machine, it's up and running. And so, you know, kids just stick the CD in there and they're off and going. When you stop running that application, it immediately gives you the option that it will take away anything it put up on the disk. So you get to decide if you're going to come back and use that often enough so you want to leave some state on your disk, or if you just want it to be cleaned off as though that application was never run.

Multimedia is another area where we've made advances. Certainly the video quality and audio quality we have today, there's a lot that can be done. But the basic infrastructure of synchronization and calling in the right codec and mixing that into the documents, compound documents, that are now quite easy to build, that's there, and it's the foundation that we can move on.

Well, next year, we'll see some other big changes. Auto-install. You won't need to think about do you have the latest driver. If you're willing, we'll just bring that down onto the machine whenever you connect up to your corporate Internet or to the broad Internet. We're doing this today in the browser with Authenticode, where it brings up a dialogue and tells you who signs the code. And, depending on who you trust, if somebody's on your trust list, then it happens without even that dialogue coming up.

Zero-Administration Windows, that's the idea of all the state on the server. And I'll say again it doesn't require new applications or new hardware; it's very, very straightforward. It's complementary to a form factor of the PC we call the Net PC, but it runs on all PCs with everything people have done today.

A place people spend a lot of money on is the directory, managing user groups, making sure that when somebody leaves that those user groups are changed, or that somebody's authorized. Today, directories don't do a good job when you want to share information across a corporate boundary, because they haven't had public key capability built in. Well, the directory that we showed at our Developers Conference just two weeks ago has those capabilities. It's out now in developers' hands and will be a standard part of the NT server during 1997.

We're also now merging the way that you find files, the way that you find messages, and the way you find pages. Those have been very different, for the way you search for something, or you say, "Has anything changed?" You had to learn different approaches. By bringing that into the shell and taking advantage of the Web metaphor, the file metaphor, and the nice things that were done in the electronic mail world, we can have something that's very, very rich there. So the Internet has really allowed us to step back and do things in a better way, connecting us to the customer in a better way.

Personally, I've been using the Internet to buy books, to learn about pregnancy and all the things that go on there -- (laughter) -- to play bridge, to stay up-to-date on the latest in biotechnology. And I've got a very nice display, a 2k by 2k resolution display, where I put up four pages at a time, and it cycles through. I'll see four news pages or four pages from competitors or four Microsoft sites that I'm following, and it'll highlight when those change.

It's interesting, though, that I'm not yet switching so that I stop reading something in print. I still read The Economist in print, The Wall Street Journal in print, the trade journals. Although I'm up on their Web sites, I'm also reading them in print. So it's just so far there's no substitution effect. I think over time you will see substitution effect.

The Internet's so popular, it's pretty amazing the mail I get all the time. Every time my e-mail address appears in an article, there's kind of a wave of those things. I have to say I feel sorry for the president of the United States, because I get copied on a lot of messages that are sent to him -- (laughter) -- and I think he probably has a tougher time getting through his e-mail even than I do.

Well, the big business story for Microsoft this last year was how we switched to really take the Internet as a top priority and got that into all of our products. And I think that's true for the industry. You know, there's almost no booth you can go to where you're not seeing rich Internet things going on. One element to this, and I think a very straightforward thing is, when you buy a PC, if you get it, you have single-button sign-up. What that means is the browser's there, the electronic mail package is there, the TCP/IP stack is there. You just click, you see the names of all the providers that you can make local phone calls to. Pick one of those, if you have a credit card, boom, you're signed up and you're off using the Internet.

We put this single-button sign-up into a revision of Windows that just went out to manufacturers a few months ago. So you'll start to see it on machines this month, and by three months from now, virtually all the machines will have this update that includes that kind of sign up.

We're also doing Active Desktop. It's kind of interesting to think what is the most important thing you want to see on your computer screen. Is it your business results or the latest news or schedule, or is the name of the folder that you happen to pick? Over the last decade, the way we use the computer screen is, we show you a bunch of folder names, you know, something that isn't dynamic, doesn't change. We should take the most important information, and bring that to the very desktop, and that's what Active Desktop is all about. Every PC will have Web server capability built in, and so this becomes the most democratic publishing medium of all time. You don't have to go out and get a specialized server. If you have a machine you're willing to turn on, leave on, the software is there, or you can transfer it up to somebody who has servers and keeps them running 24 hours a day.

The security software file wall and proxy server software, big advances took place this year. It's very easy now to configure on NT and other systems all the safety and sharing that you want. And we're getting now video streaming and audio streaming as a standard server feature.

The Internet is everywhere. I mean, you just can't get away from it. The URLs, the demand for bandwidth, there's even a social phenomena now where kids sort of compete to see who's got the best connection. And we did a little video to update everyone on some of these wild things going on. So let's take a look at the latest online mania.

(Videotape presentation.)

MR. GATES: We have a lot of fun stuff going on with the Internet, but the Internet is also serious business. And I think this is the year that productivity applications and the Internet will really come together. This is very exciting because we take these applications that have been built as components, but more than components, components that were tested together with the unified interface, and bring those onto to the Web. We use the Web to make it so that you can upgrade very easily, and we use the Web metaphors so as you navigate around in compound documents, or even to HTML pages, you've got your favorites, your history list, that linking metaphor works very, very well there.

But the Intranet is more than read-only. It's not just looking at pages and reading them, it's also creating pages. It's annotating pages. It's editing what other people have done. And the top problems that productivity software has been addressing, letting people view data in different ways, letting them coordinate their activities, all of those things play very well for the Intranet. We basically moved from a file sharing world, where information is just too hard to find, into a world of pages. Pages that can be the Office-type format.

A very important change is taking place here. The way that we relate to a productivity customer will be completely different. Instead of thinking about buying a box every two years, instead you'll have a constant connection. If you've given us permission to send you mail, say, once a month, we'll have some profile bits that come up about your use of the application and your hardware, and so the mail we send you will be customized to your usage pattern. We'll tell you about more clip art we've got on our Web site, new templates, usage hints, some updates that we might have available, or even new components that you can just click and have those come down. So there won't be this big one-time event.

If we notice in your profiles events that you've gotten more hardware, then we will suggest new functionalities, but we'll never suggest things that go beyond what you have available and what you can just automatically go and install and fit with your use. So it's a much more customized connection. And, as I mentioned earlier, we use that for feedback across the product. And so it's a very different for how productivity applications relate to their suppliers.

There are a lot of challenges with the Internet. You know, people can view this as a glass half full. And there's been so many wonderful things written about the Internet, there's no doubt we'll see articles that are kind of a backlash saying, well, is it really all that people said it would be? Well, in the next two or three years, there will still be shortcomings, but I think to really understand this thing, you have to think out 10 or 20 years when a broad set of people will see using the Internet to get information as part of their daily activity, and they'll expect everything they do, whether it's scheduling a doctor's appointment or negotiating a contract, or trying to decide on a purchase decision, they'll use the Internet as a tool for that.

Bandwidth is one of the tricker problems. At one of the industry conferences, a speech was given saying that backbones do not obey Moore's Law, and this person has really researched backbones and pointed out that backbones were improving about 2 percent a year. And so that's very nice, I didn't know they were improving at all. But it's not enough to expect there to be optic fiber everywhere overnight. In fact, the best thing that happened to bandwidth in this last year is people like Rockwell and U.S. Robotics figuring out that they could use the normal phone infrastructure and double the speed up to 56k baud. That's a fantastic step forward, and it will be very pervasive. ADSL, ISDN, PC cable modems, those are great things, and we're going to do everything we can to help raise the adoption rate of those things, but it's going to take a long time.

The Web does need somewhat richer structure. If you've seen a schedule on the Web, you ought to be able to take something like a concert and just drag it to your local schedule and have it show up. Well, that requires using object technology inside the pages. So, another great place where we can evolve object support in. We also need to take links and give them types, so that if I go to a site and say I want to use this site offline, it can understand based on the link types what pages to bring down. And so we need new standards in that area. The standards committees are working very well here -- W3C, IETF -- and so I think this richer structure will come quite rapidly.

We need to get security in place and get people to trust that. Good progress here where the credit card companies have gotten involved. The only tough problem is the authentication, knowing who's at the other end of the line if they're doing something that's hard to reverse.

A final issue, and one that's more of a societal challenge, is, given how great these tools are, can we make them available to everyone? Well, not everyone is going to be able to have one at home, so we should use schools and libraries to fill in that role. And this year, I've gone out to many libraries of the hundreds, where we've given grants to have PCs and Internet connections, and it's fun to see that they really do get used. A new group of people comes into the library and thinks of it in the same way it was thought of 100 years ago, where books were the key new tool, and the library was the place that you could come and get the same book that other people had available.

Well, let's think back to the original vision video that we saw. There was a lot of neat stuff there, very high-reach stuff, hard core R&D, and some of these are things that people have become a little cynical about. You know, will computers ever understand speech? You see a great demo, and then somehow it doesn't work, and so people don't understand what's really needed there. Well, the problem is, without context, it's a very, very tough, tough thing to do. Even humans couldn't recognize speech unless they knew what the context was. And so we have to tie together rich elements in the system.

We're hard at work on these problems. Of our billions in R&D a substantial portion is going into a high-reach effort. And, in fact, these are the things that will use those incredibly high speed processors that Andy Grove and I talked about -- Andy Grove talked about yesterday. Let's real quickly see one last little video, which is a tour of the lab that Microsoft has, and a quick glimpse at some of the projects going on there.

(Videotape presentation.)

MR. GATES: One of the key points is that just doing a rich input system is not enough. You'll have to evangelize to every application developer how to describe their context. And the way that context is described is actually a very, very key piece of technology. So these input systems -- handwriting, speech, visual recognition -- all of them need that context capability.

Sometimes when I talk about visual recognition, people sort of think, "Well, that's a strange thing, the computer is actually going to know when I sit down." It's going to know how you're reacting to things. In fact, if you're willing, we could define a new type of return receipt mail that not only says that a piece of mail was read, but also described the reaction that somebody had to that piece of mail. (Laughter.)

These problems do require a lot of computing power. And what I heard yesterday was very good news because more RAM and processing cycles are fundamental to being able to do this well, particularly this idea of learning. That's been discussed as an agent technology that models what you are after. But it's a very, very deep problem.

Interestingly enough, the kind of research I am talking about here, the total funding of these activities is probably less today than it was 10 years ago when people were somewhat over-optimistic. And yet I feel that we are definitely within reach of making these things reality. This is what I mean: making the personal computer more personal. And that means having more cycles, but hiding what's going on rather than going back and having less cycles.

The end goal here really comes back to the vision of information at your fingertips. The PC is the ultimate empowering tool. And although first we'll see how that's used in business, and second we'll see how it's used in homes, I also think that having it universally available in libraries and schools is very, very important.

I'm actually quite envious of kids growing up today. They won't have to go through all the things we have gone through. And, in fact, when we look back at today's personal computers, I think we'll say, "Hey, these were the machines that couldn't listen, couldn't talk, couldn't see." The operating system, yes, it will still have a file manager and a multi-tasking manager in it, but 90 percent of the code will relate to these new input systems. And so the tools here that these kids will have to pursue their curiosity and do new things, will be quite amazing.

I'd really summarize by saying I think we're all lucky to be working in this field. We're providing tools that empower people to do their best.

Thank you. (Applause.)