Remarks by Bill Gates
Microsoft Corporation
Friday, September 26, 1997
San Diego, CA

MR. GATES: Good afternoon. It's been very exciting to see the response to the PDC. I know in the weeks building up to this, we were really pleased that we were going to be able to get the Windows NT 5 beta out, and talk about all the things we're doing with COM+. So we've used this PDC as a really major milestone, interms of our message to developers, about how to build modern applications.

I was looking forward to coming down for the party last night. We didn't anticipate there would be a hurricane. But, I guess it has a happy ending. We took all the food that you were supposed to have, and the night's entertainment, and we sent it over to a homeless shelter, and the YMCA, and they made very good use of it. And many of you got to come to late-night sessions about COM, instead of going to a party. So everybody got what they wanted.

When putting this speech together, I'm thinking, boy, in the last 12 months I've given about 100 speeches. And I'm very careful to try never to say the same thing twice. So I was looking back through some of those video tapes thinking, well, maybe I could take a few gems, and just use those. And as I was going through it I thought, well, maybe I'll just make a short little video tape, and show a few of the highlights of how I'm careful not to go over the same material. So let's just take a quick look at how that's done.

[Video spoofing things Bill tends to say in every speech.]

Well, it is a very, very exciting time. If we look at the piece of technology improvement, I'd say it's even faster today than in the past. One place we always look first for that is the processor itself. The improvements in clock speed, the improvements in the amount of work that's getting done on a clock cycle. Intel is leading the way, putting billions of dollars into research. And they continue to make very rapid advances. We're working with Intel not only on advanced Pentium processors, but also on their Merced chip, making sure that NT will be there with full 64-bit support the day Merced comes out.

I think the other elements of the system are also carrying their weight. The drop in the price of memory. The size of the hard disk. In fact, with hard disks now being several gigabytes, we can look at schemes that save substantially on communications costs by doing broadcast data and caching on the local machine. You know, the storage is much greater than the user is ever going to type during their entire lifetime. And things like the replication approach and IntelliMirror, which is a breakthrough in terms of usability, roaming and in every other way, those are very simple to do now that storage has become quite inexpensive.

In the years ahead people are underestimating the advances that will take place in screen technology. I've got a nice 20-inch LCD on my desk, and I love it. Now, it's not too cheap yet, it's about $6,000. And my house has a lot of those in it, so it does kind of add up a little bit. But that kind of screen display is going to be available in high volumes at reasonable prices sometime in the next five years. And so this tradeoff of what you read on the screen versus what you read in print will start to move more and more towards the screen. In parallel, the way you distribute information, even if you're going to print it with low-cost, local printers, will also move in the same direction.

PC volumes continue to go up at a pretty astounding rate. It's at over 80 million machines a year. And as the low-cost PC moves from $1,000 down to $800, down to $600, we'll be opening up new markets, growing the home market, letting people have multiple PCs, and so the future here is very bright. It's a foundation that we can all build wonderful software on top of. Even for the most difficult problem, which is the high-speed connectivity, we can see progress. Microsoft is investing in ADSL and cable modems to really drive both of those forward, and get a good competitive dynamic so that both cable companies and phone companies are making the right investments there.

On the side of business, I think that we'll have very good connectivity within the next three to five years, even moving down into small businesses. In the home, it's going to take longer, and that's why very clever approaches using local storage will have to help us get there.

In the PDC, there are a lot of messages, and this is my favorite slide from the PDC. It sort of tries to put it all in one place. It captures the idea that we think mobile use is very important. It captures the idea we want to still give people the benefits of centralized storage, and that's where IntelliMirror comes in as part of our zero admin initiative.

Windows NT 5 is the breakthrough not only in the directory, but also in the administration capability. And the other dualities, the best of pages, the best of codes, the best of Web server, the best of client server, and finally in the storage area, the best of structured and unstructured storage. This is the direction we're going.

It's fair to say Microsoft has bet its future on Windows NT Version 5. We'll be driving the business market to use that product as rapidly as possible, and then we'll do a variation using the same technology that will drive into the consumer market. And so, sometime in not too many years, new personal computers will come with the refinement of Windows NT Version 5 in the same way they come with Windows 95 today.

We've laid out the DNA framework, and we think it's important to give people that direction to really show how we're filling all the pieces in. Certainly, it's built around the things we have today. It's not revolutionary, but there are some big advances. COM+ is a huge advance in terms of the accessibility and ease of getting at the power of COM, and all those services that we're building around it, whether it's transaction management or load management, all of those are going to make it easier to write very powerful applications that scale and are usable in a very distributed fashion.

So the themes that you should take away from this conference are: the Windows NT 5 commitment as the platform for Microsoft; Windows DNA as a blueprint for building the applications; and COM+ as the way the plumbing will get done, whether it's within a process, whether it's across multiple processes in the system, or whether it's between different computers. And we have our entire development budget focused on making those really great, and creating enough volume of every one of those things that that's where you'll see the demand from the software market.

We spend a lot of time with large customers, and it's interesting what they're saying to us now about what they'd like to see out of their computer systems. They understand that they've invested a lot in having PCs and networks for simple productivity applications. But now they're saying, given that investment, can we manage it better, and can we start to use it for the fundamental information processes within the company? And we use the term digital nervous system to refer to this. Nervous system means the way that information flows in the company, encompassing everything, meetings, paperwork, fax, phone calls. And digital nervous system is the part where people are taking advantage of the connected PC to do the information exchange in a richer fashion.

We believe the companies that step back and really think through their key processes and put those into electronic form will be the ones that get ahead in the future. Part of it is very mundane, just taking paper forms and getting them on the machine. Parts of it require deeper thinking. How should a company respond when a competitor does something clever? How do you make sure that the bad news spreads very well, and people collaborate together to come up with a quick response plan? When a project needs to change, how do you make sure that people in different locations are not only aware of that, but are working together very quickly to come up with a new plan?

So a digital nervous system starts with simple electronic mail. It starts with the intranet, but it's also about thinking through how you reach out to the outside world as well. It's not just about planned events like business planning or personnel review, but also about unplanned events that are very key. The ingredients are very simple. A PC, a network, and hopefully lots of software from all of us here. I think that is the key element that will take all the power that's inherent here and really bring it to the fore.

And one of the things we'll be doing is really highlighting companies that do this well. So, they'll be the pioneers. And other companies will look at that and decide that they, too, want to participate in this. And they're already spending most of the money they need to do it. In fact, by making the manageability better, and by doing things to reduce data center costs and communications costs, and development costs--this will allow them to put more of that budget into buying packaged software and customizing that software. Without a huge increase in spending, they're getting far more strategic leverage out of the money they're putting into IT.

I want to look a little forward here. You know, the NT 5 thrust, the DNA thrust, the COM thrust, those are going to be the big things rolling out in the industry over the next three years. But during that time, we'll also see the beginning of a change in the way we interact with the computer, moving from simply having the keyboard and the mouse as the only way that we get information into the machine. Our belief in this is partly why we've funded a very good science research group. Overall R&D is over $2 billion, and a growing portion of that are activities that really deserve the label of pure research. We do this in many locations: Redmond. We just opened up in the UK with a great group there. In San Francisco. We'll be opening up several additional centers. We do it in cooperation. We're funding a lot of university research, which is where, in the world at large, I'd say, the majority of great research is done in the universities, and that's a great foundation to work with them and to build on the great ideas that come out of that.

It's important that lots of papers get published, whether it's on graphics, or linguistics, or new operating system approaches, and we definitely participate in doing that. The synthesis that we're looking for here is a far more intelligent computer system. Something that's easy to use for common scenarios, and that you can interact with very easily.

What are the key areas? Well, certainly databases to get arbitrary scalability; new programming techniques, one that we call intentional programming, that's headed up by Charles Simonyi, who was one of the first people who did some of the great work at Xerox Palo Alto Research Lab. Program analysis: we're going to change the way that programs get optimized from simply being compile time and run time to having different intermediate stages, so that as soon as you know the specific processor you're running against, you can do optimization there. And by having very high level descriptions of the execution paths, we'll be able to do optimization at more than just compile time.

The things that are easiest for the world at large to appreciate is what we're doing with speech and vision, because those bring up a completely different image of what the computer will be. In fact, the image that's been painted for many decades in science fiction. Now, 20 years ago, people were very optimistic that these would be easy problems to solve. And the demonstrations looked very good. In fact, these are very tough areas, and it's only by having all the great power of the exponential improvement we get through Moore’ Law, along with a lot of very rich software, that these problems will be solved. And they won't be solved in isolation. You can't just do speech recognition without linguistics. There isn't enough information there.

The way that humans disambiguate what they're hearing is by having not only common sense about grammar and words, but also the deepest context about exactly what's going on at that time. And so the way that applications interact with these new natural input systems will require a lot of rich exchange of information so that the high probability inputs are the ones that get chosen.

Well, first, let's look at text-to-speech. Of all the things here, I think this is the one you'll see in widespread use most rapidly. Current processors allow us to do very reasonable text-to-speech conversion. [Video.] So, there, the computer is starting to sound fairly reasonable. Part of this technology is to be able to take what we call a voiceprint, where we listen to somebody talk, and capture in a few hundred Kbytes, the essence of what their voice sounds like, and then being able to apply that to different texts.

I've got the little utility here that will probably put out on the Web when it’s available. It's the Microsoft Mouth, it sounds a little bit like the Microsoft Mouse, and that really makes sense, because we talked about putting the microphone in the mouse itself. So it would be a combination mouth-mouse product. But, we haven't been able to make the business case for that yet.

So I can just say anything I want, you know, “There was a very, very big hurricane in San Diego.” Let's see how it does.

[Computer’s oral output echoes Bill’s oral input.]

Computer: There was a very, very big hurricane in San Diego.

MR. GATES: Okay. Let's try an error message. The computer is sorry it can't help you.

Computer: The computer is sorry it can't help you.

MR. GATES: That ought to please people. And so there is no extra hardware here at all. It's just simply using a standard Pentium processor. Now, when you combine speech synthesis with speech recognition, that's where you take the biggest step forward, because you can just, say, call in on the phone and get your electronic mail. And there are some start-up companies and people like IBM and Dragon who are doing this.

We see our role as doing the basic technology, and integrating it into the operating system in a way that people can do applications that use these techniques. One of the other things people look at is synthetic singing. I'm not sure what the application is, but this should give you a sense of the advance that's taken place in the last 25 years. Here's 1962 Daisy, which was the inspiration for Hal in 2001.

Computer: Daisy, daisy, give me your -- (inaudible) --

MR. GATES: Not too good. Here is a little more modern singing.

Computer: Are you going to Scarborough fair, parsley --

MR. GATES: I don't think singers have to worry too badly, but another 25 years. … Realistically, where are we? Humans are amazing at doing speech recognition. I mean, it's almost hard to believe how well they do. If you just have digits that are being read off, humans have an extremely low error rate, about 0.01 percent. And the machine is still at 0.7 percent. That still means it's reasonable in some applications. But, you know, that's a huge gap. Continuous speech dictation, which would be either our technology or some of the others, you have about an 8 to 10 percent error rate, humans 0.4 percent.

And conversational speech is the most difficult. Humans are so good at having the clues, and dealing with the missing pieces that they manage to get only a 4 percent error rate, despite the fragmentation. Computers are still fairly poor at this. And so we have more work to do. We need more MIPS, we need more memory, and we also need the linguistic system and the speech system working together.

We actually started first doing the linguistics. And linguistics is also making rapid progress. Simply recognizing a spelling error, that's standard stuff. Recognizing a grammar error, that's now standard shipping capability. The next thing is a case where, to the computer it seems reasonable, but the way a person would parse this would suggest that it reorder the words so it doesn't sound strange. And the ability to do this right is what we have at the research level now. We have a thing called Mind Net that describes for every word, basically, how it's used, how it's used with other words. And it's only by having a rich dictionary that talks about the world itself, that it allows you to recognize that there is something definitely wrong there.

The next level of richness that we don't have yet, and no one has, is the ability to do translation. Here's an example in a Norwegian bar, I think warning about the dangers of alcohol. But, somehow it didn't get translated in exactly the right way. To be able to do that right, you have to have a very rich world model, not just about the individual words themselves, but reasoning about what the intent of the sentence in English meant, because if you take the words in isolation it does translate into obviously the wrong semantics there. And so each system will have to improve a great deal.

Now, both the capability of having speech synthesis and the ability of doing natural language queries are pretty much here today. We just shipped a version of SQL Server, the Enterprise Edition, and we included in natural language technology. This work actually started over eight years ago, in a little company called NLI. And when we saw their technology, we bought the company. The great people from that company came in and helped us integrate it in with the database.

So here I've got SQL Server connected up to a movie database. And when you do the connection, you need to do a little bit of work to describe the standard vocabulary, so that the sentences can be turned into SQL queries. It took about two days to do the vocabulary for this movie database that's called Cinemania. So we can say something like, how many movies did Michael Caine act in? And I have to be smart enough to recognize that Michael Caine is essentially a data constant. Now, it restates -- it thinks that the word perform instead of act, it knows that's a synonym, and then it shows us the SQL, sort of reminds us how obscure SQL is. And then it tells us that Michael Caine is pretty prolific. He’s acted in 82 movies. I used this earlier to find out he was the person who acted in the most movies of anybody.

We can take something a little more complicated: Show me all the movies that were given five stars. I didn't ask who gave them five stars, so you might disagree with some of the ones in this list here. In fact, there's a few that I don't think are very good at all. But, that's a very complicated thing for the computer to help us with. And then just one more, what roles has Harrison Ford played after 1980?

So these things would be fairly difficult to do if you had to do the SQL yourself, or do the navigation yourself. But, they're very easy to do. And so we see that Harrison Ford has played a lot of different characters, in a lot of different movies. So on a Web site, imagine connecting this up to a travel database, a shopping database, or within a company, connecting up to do sales analysis. So it's very straightforward, and people can just walk up to it and expect to use arbitrary sentences in order to work with that. So when we combine that with speech recognition, you've got something really pretty dramatic.

Another place we'll be using linguistics is in this Web search area. Today, great search engines like Alta Vista find you a lot of stuff. In fact, they find you too much stuff. If you do this query, who sells the fastest chip? You'll find out about potato chips, and fish and chips, Wall Street blue chips fall. … And you'll find one that actually is relevant.What's being done right now with the search engines is, they’re just taking those individual words and doing a little bit of proximity analysis. It's not parsing the pages to see how the words are used. So the right way to do it is to take all the sentences out there on the Web, and figure out the linguistic structure. So when you say, who sells the fastest chips, that is assuming a relationship where the chips are fast. And so when we actually apply this to the Web, we went from getting about a 33 percent set of articles that a human looking at them said, yes, that really had something to do with it, going up to 63 percent. And so that's a big increase in the number of relevant documents that you can get using natural language.

It's very important for us to start to sort through this information overload that we're getting out there. It's my belief that every computer will have speech and linguistics built into it, because that's the world of discourse that we live in. I'm not saying the keyboard will go away, but a substantial part of the operating system will be this ability. And it will use up a lot of local computation. You won't be shipping your speech waves up to a server and doing time sharing with a lot of latency to have it come back and do the recognition. You'll be using the local power and so we really need people like Intel to keep making the investment and keep Moore’s Law true, for at least another 10 years, before we'll have power to spare for these very key tasks.

Another thing I'd mention, which I'd say is at definitely an earlier stage, but is very exciting, is the idea of computer vision. Certainly, low-cost video conferencing cameras, for about $100, are being included into lots of new computers. And so if you just put a tiny piece of software -- not tiny, a very nice piece of software behind that, you could start to do some interesting things. And I asked Matthew Turk, who is with our vision research group, to come on down to San Diego and show us a little bit of the progress that his group is making in this area.

MR. TURK: Our group is working on computer vision, to try to give computers the ability to see, so that we can interact with computers not just through these devices, mouse and keyboard, but also through visual cues. And this will be very useful in the short-term for several kinds of applications, people with disabilities, for games, social interfaces, et cetera. I just want to show you a few little demos of these abilities.

[The computer recognizes that Mr. Turk is now in front of it; the screen comes on.]

MR. TURK: Here's the guts behind it. What you see on the far left here is a background image that it took when I left, and it saw there was no movement going on. Then you see the live video of me moving in front of the camera. This white and black thing is basically the difference between those two previous images. So when it sees something new, it thinks that's probably a person, and it encodes that as white. And then on the far right, it’s basically a draping over what's there. Okay. So it's a simple representation of the stuff in front of the computer.

Now, based on that, I can then go back into the original image and find where my head is, so you can see it's tracking me, the green box moves as I move around. And then based on that location I can do some simple gestures, like --

[As Mr. Turk moves his head side to side or up and down the computer responds: No, no, no, no, yes, yes, yes, yes.]

MR. TURK: I wasn't touching anything. I was looking for yes and no there, shaking my head. I can also, at this point manually, save some particular instances of what I look like. So I'm going to save one here called head position, when I'm just here standing at the computer. Let's make a simple gesture, like salute. I'm going to salute with my right hand, I'll save that. I'll salute with my left hand, I'll save that. And I'll go home, and save that.

Now that I've saved these things, I can recognize them in real time. So the red that you see is what's being recognized at the current moment. So right now it says that I'm here. If I salute with my left hand, that lights up. Salute with my right hand, that lights up. Let me turn off the display, just to speed this up a bit. Okay. So salute left, right, left, right, get out of here. Okay. So you can see how this could be the basis of a game of “Simon Says” or something like that, or in the case of the screen saver, it was basically looking for head position. And when it saw head position, that's when it turned the screen saver off.

Okay. Now, I can use the same substrates to play a little game. In this case, a game of tic-tac-toe. So the first thing I'm going to do is to train the system, as I just move my head around here. And when it's finished training, we're going to play the game, without touching the mouse or keyboard. Okay. So you see the red box is the feedback, where it thinks I'm looking. And I'm just moving around here. When I stop for about a half a second, it plays my X. Okay. Then the machine plays an O, and now it's my turn. So I'm going to go down and play an X down here. The machine plays its O. I'm going to go play an X down here, if I can -- oops. Oh no, I lost.

MR. GATES: I've never seen you lose before.

MR. TURK: Well, I programmed this, too. You would think I would win.

Okay. Let me move on to another completely different demo. There is actually -- this one is using this camera here, which is actually a digital camera. It's the first time I've actually used one of these. And the idea behind this little demo here is that we're building an environment for children, hopefully, to interact with computers, particularly young children, who are too young to do anything interesting with a keyboard, but they just want to do things in front of the machine and have it react to them. And music is a really nice environment for this.

So the first thing I'm going to do here is show it recognizing clapping. I'm just going to stand here in front of the machine and clap my hands. Okay. And you hear a cymbal sound. And next I'll play the bongo drums by doing a different gesture. What you're seeing here on the very left is the live video, the next image is basically what has changed from the previous image, so moving things. And then the far two images are encoded left, right, up, down velocity. And then flapping wings, let's say a kid's trying to be a bird.

And then finally I'm going to ask Bill to direct a song. And to come stand over here. Okay. Bill has always wanted to be a director, so let's see if you can do this. Okay. And finally, let's see how this one works. Thank you very much.

MR. GATES: So that's where vision is. And at first you'll see it in games and educational software. But, eventually it's a fundamental part of the computer interface. So definitely the PC is not easy enough to use. There are some simple things we can do to make it a lot easier, things like auto install, auto repair, IntelliMirror, where you don't have to worry about your state always being backed up, or if you want to get a new machine, having all your state replicated to that. Those are important things, but to really have people think of the connected PC and all that information as part of their every day activity, where that's how they go and decide where to travel, what to buy, how to stay in touch with people, I'm going to have to have a seeing, listening and learning computer.

And even though Windows has a lot of functionality, this new code base on the client side will be the majority of the code that's there. Now, we need to take the right time frame here. It's not going to happen over night. But, I will stake my reputation on the fact that within the next decade this will be the centerpiece of Windows on the client. We need Moore’s Law in order for that to happen. And we're going to need to work with people who do applications and show them how to bind into these interfaces, so that no matter what people are trying to get done they can use the natural interface.

And so I see, in the near-term, a wonderful opportunity for developers. Looking out into the future, I see even more opportunity. And so we look forward to working with you to seize those opportunities. Thank you.

QUESTION AND ANSWER SESSION

Question about encryption.

MR. GATES: That's a very timely question. I spent a fair portion of this week on the phone talking to people in Congress about what a bad thing it would be in many ways if the new legislation that some people are considering were passed. To date, we've had a huge problem, because we can't export decent encryption. We're stuck at 40 bits, which is easy to break. And the foreign software competitors are shipping lots of 128-bit software. And so our customers outside the United States say, hey, you are providing us inferior products. And we have to say, well, it would be a felony to provide you with decent encryption.

And so we went to the Congress and said, come on, this is backwards looking. Good encryption is available to everyone. And you should allow mass-market software to include the encryption technology that's widely available throughout the world. That's where it got started.

But, what happened is, some of the domestic law enforcement people, who we don't think have spent as much time looking at the technology and where things are, they came in and said, no, instead of freeing up the export, we're going to restrict domestic use of encryption. And we're going to insist that every software product be restructured so that there is a back door, so the government can get in and see exactly what you're saying, without you knowing about it.

And that's literally the bill that's being considered. By '99 we'd have to break the software we've got out there, and never ship anything that doesn't have this simple back door for the government. And it really doesn't address the fact that people who want encryption, and use it to hide bad activities, will still be able to do that. So I think this is a serious issue that hasn't gotten as much attention as it deserves.

It's not just a business issue for everybody in the software industry. It really is an issue of embracing the information age, which other countries have done. And it's an issue about civil liberties. So I hope there is a lot more written and discussed about this. We're certainly going to do our part to make sure that the right flexibility, the right freedom is there.

Question about Java and HTML.

MR. GATES: Well, we believe Active Controls are the way to get the most flexibility, in terms of having code run on the client. You've got really three levels. You've got HTML, including DHTML, you've got the scripting capability, which I think most of what we showed here wasn't Java, a lot of what we showed was various types of scripting, either JavaScript or VB script. And it's important -- there's a big, big difference between that and Java. And then finally, if you want to write in any language, we have the Active Controls.

We do think that the biggest component software business out there today is Active Controls. And those are written around the COM standard. There is going to continue to be a lot of innovation by us and other people there. And so if it didn't get the right emphasis, it's almost like we take our success there for granted. Internet Explorer is getting higher and higher market share, to be able to host more of those controls. And the third party add-ons that let those controls run inside Navigator have improved quite a bit. So we think the momentum there will continue to grow. And that, in a wide area sense, most people will stick with DHTML and scripting, as opposed to either Java or very, very rich controls.

Question about Teledesic, which plans to send up low-orbit satellites.

MR. GATES: Yes, I'm an investor in one of several companies that plan to put up satellites, at about 500 miles high. And part of the beauty of being at that height is you get low latency, and the ground station is quite inexpensive. You can have a several hundred dollar, both send and receive ground station. Now, the drawback of putting the satellite at that height is that it won't stand still. There is no way to make it geostationary. And so, in order to cover any part of the globe 24 hours a day, you literally have to cover the entire globe, which in the case of the Teledesic design, it's 288 satellites.

Now, where does Teledesic fit in? In urban areas, where you have a lot of density, Teledesic doesn't add much value, because in a square mile it can only connect up about 100 people. It's only when you get outside the city, and it's not economic to run the fiber or the other improved infrastructure, that's where Teledesic fits in. So it's very complementary for rural areas, anything that doesn't have high density, it will be the low cost solution. In the cities, we're going to have to believe in cable modems, ADSL and fiber. And so Microsoft is doing everything we can to encourage that.

It looks like ADSL will essentially replace ISDN as the broadly deployed, high bandwidth solution. It's a more modern approach. It's got better data rates, and the companies doing the chips are going to be driving those prices down. We've got about 1,000 employees connected up at 6 megabit ADSL, and it worked very, very well in the trial. And so we're getting phone companies all around the world to roll that out.

But, I would say you still need to think of more than a five-year time frame, before you can say even small business customers as a very high percentage will have that kind of connection. And so the idea of using broadcast data or being clever about using -- still using the narrow pipes, dial-up pipes, that's going to be important to all of us for quite a bit of time to come.

Question about Bill’s programming.

MR. GATES: I wish I got a chance to write more code. I do mess around. They don't let my code go in shipping products. They haven't done that for about eight years now. And when I say I'm going to come in and write this over the weekend, they don't really believe me quite as much as they used to. But, certainly when it comes to the new things we're doing, you know, I always thought COM, the source code, was a little bit verbose, even you might say arcane or ugly. But, COM+, to me, I'm very excited about that. I mean, it really is the sort of direct access, using these component libraries in the right way. And certainly I'll be able to clean up some of the things I've hacked around with.

One of my favorite programs is one I did on my nice little 20-inch LCD. It brings up Web sites. And so it will bring up four at a time. So, it will bring up like four competitor sites, or four general news sites, or four trade magazines. And it just rotates through. And then it notices if I click on something, it makes it full screen, and it makes it easy for me to change things. That's just actually a nice little VB application that one other guy and I hacked together. So I play around with that quite a bit. And it kind of gives me a sense that there are still some things that are fairly hard to program. I even found one or two bugs in the process.

MR. NIELSEN: Those bugs got fixed.

MR. GATES: They did. Good point.

Question about Sun’s position on standards and Microsoft’s position on standards.

MR. GATES: You know, our industry relies very much on standards. And Microsoft and many other companies put a lot of time onto the standards groups, W3C, IETF, ITU, many great groups. When something is called the standard, what that means to me is that no company has a privileged position. That it's a process where anyone's good ideas can come in, nobody can arbitrarily change the criteria or control the trademarks. So that's standards, and that's great. And I'm not saying that's the only thing. Then there are proprietary products that some companies invested a lot in building, and they incorporate standards. And Windows incorporates, literally, over a 100 of those standards, but we don't just put the source code out. It is a product of Microsoft, taking input from people, but we set the direction.

I think to try and confuse and say that, yes, Sun controls the tests and the brand name, and they decide what everything is, and yet to call it a standard is just sort of an abuse of the standards process. And I think it's great if they decide one way or the other. Either Sun should give up having a unique position, just be one of many companies who are contributing to it, and let somebody who doesn't have Sun's commercial interests, a neutral body, control it. Or, they should just say, hey, this is something we control. It's not a standard. And the letter that several companies sent them a few weeks ago really sent them that message, and they're still trying to have their cake and eat it, too, say it's a standard, but really Sun has a very unique position in terms of everything that counts.

Question about cross-platform support for COM.

MR. GATES: The evolution of COM on a cross-platform basis is something that we care a lot about, and we've really strengthened our relationship with Software AG in order so that people know Microsoft stands behind the COM capability on different platforms. And it's up to software developers to tell us which parts of COM, COM+, they'd like to see on a cross-platform basis. What we have today is the key thing, we have deep plumbing so you can create your objects and call your objects in the standard way, and there are a lot of developers who have used that very successfully. If we need to evolve that, put more of the elements in, you know, we're very open-minded to it, but we see that as part of an ongoing dialog.

Some things, like SQL Server, or Transaction Server, we won't port to other environments. But having the same interfaces, standardized interfaces, so that whoever does the transaction management on those other systems can plug into those. That's something that we think is just fine.

The standards group that's involved there is the one who will officially ask what's going on the cross-platform COM. I think whatever we get from input from developers is what we'll do in terms of extending it there. Today, it is just COM, it's not all the extra things that were described here. And so that level is something that will depend on what the demand is.

Question about hiring good people.

MR. GATES: I think hiring people is one of the toughest problems we all face. Certainly, at Microsoft from the very beginning of the company 23 years ago, the key bottleneck has never been financial issues, but rather how quickly can be bring in more great people. Now, we've gotten better at that every year. You know, we're hiring well over 1,000 new developers, most of them out of college, every year. And, you know, we do a lot to make it an attractive place to work. The industry as a whole faces a shortage. And I think the market mechanism will work. People will see that these are interesting jobs, attractive jobs, jobs that are going to make a difference in how things work in the future, and that there will be a rise in supply. And so, I don't think in the long run there's a problem. But for the next several years, it's a challenge.

We even find ourselves going back to the Congress and making sure they don't do things to really horribly restrict immigration. We want to allow very talented software people from outside the United States who want to come and work here, allow them to come work in the country. And there was a close call about a year ago where all of that was going to be cut off completely. But, you know, hiring, that's a big issue.

I don't think it will lead to higher software prices, honestly. The key to software is selling it in high volumes. You know, that's the way we spend $2 billion, and yet sell an operating system for literally hundreds of dollars. We just sell a lot of copies. And I think that's always going to be the key to allowing the industry to invest more in R&D and yet pricing for a very, very large market.

Question about whether the information age will improve politics.

MR. GATES: You'd find me an optimist. And I think people are interested in being informed, and it's a little hard, you know, with short articles and sound bites to take any topic--what is the budget really spent on, or what did my representative vote for, who else cares about this cause that I could talk to and we could organize to do something about it. The Internet gives us a way of exploring topics in more depth, having things presented to us, and really showing us, based on having your Zip Code, say, in a profile, for every article you read, you can see what your representative voted, or did they make any particular comment.

And so I hope that it can draw people in and allow them to care more about political issues. I mean, that's at the center of democracy, that voters care enough to think about how their casting their votes. And so, like most things about information technology, I think people will take advantage of that. Certainly, if you consider the generation coming along that will grow up with this technology and take it for granted, I think they'll do amazing things with it.

Question about Microsoft support for Java.

MR. GATES: Okay. When people use the word Java, they can mean two different things. They can mean Java the language, which we think is a good language and we believe we've got a great implementation of that language. Certainly, it is the most popular set of tools for that language with debugging and all the things we've done there. And so, we feel good about that.

The second thing they can mean is that instead of taking advantage of the computer that somebody owns, you just use the vanilla set of run-time services. So, if I own a computer, whatever it is, a Macintosh, Windows computer, when you're shipping that application, what you're saying is, we do not use the user interface, or the clipboard, or the color management, or the high-speed graphics. We use none of those features because we've written an application. We're just kind of lazy, basically, and even for the environment that's 90 percent of the users, we're not going to exploit that. And so that's the question, it's for rich applications, do you ever want to take advantage of an operating system, or take advantage of a database, or a middleware product, and certainly it's our view that people are going to keep calling pieces of software. They're going to call Oracle. They're going to call Notes. They're going to call Windows NT. They're going to call transaction managers. And so this notion that you can write something that runs on a wristwatch, and you say, wow, this is the best spreadsheet I've seen, it runs on a wristwatch. Thank goodness that, you know, I've got this big screen, but it's only using that little part of my screen there.

Anyway, it's almost strange in a way to think of not taking advantage of the customer's environment. And so it's when you talk about the “purity” aspect of non-exploitation, there Microsoft just says, okay, we'll see what software customers want and what goes on there. And, meanwhile, we're doing a lot of innovations to give people a choice of languages. COM+ in some ways goes beyond what any language has delivered, the transaction management, the load balancing. In some ways, it brings two languages like C and Visual Basic, the garbage collection that has been a natural part of what's there in Java. And so people can take C code forward, they can mix in Java code.

So, you'll find us very agnostic about languages. We're going to continue to evolve Visual Basic. We're going to continue to evolve Visual C. We think other languages that we're not involved in, like Powersoft or Delphi, or a dozen others that I'm not mentioning, all of those, that there will continue to be code there.

It's very hard to go to a CIO and say, it's your lucky day, rewrite all your applications. They'll only be a little bit slower. You know, it's tough. When are they going to get it done, and why?

[Applause.]

MR. GATES: So, Java is fine. But, you know, be careful at the multiple definitions that somebody may imply when they use that.

Question about the quality of the documentation of Microsoft products.

[Applause.]

MR. GATES: I think it's fair to say that with our Internet products, the first few versions that we put out, the key to us was having the functionality and having very high performance. And so even though I think they were very high quality products, we didn't put an emphasis on making it easy for the developers to get at those products in coming up with great documentation. We wanted to have the fastest HTTP server, and so we got that out there. And, fortunately, we got ASPs in there, but we didn't tell people a lot on how to take advantage of them.

Now, you're seeing the emphasis shift. So, both in terms of examples that we have out, and doing more material, I think you'll find that there really has been a shift there. We do put our TechNet information up on the Web, and we keep that up-to-date. And people who belong to MSDN have access to that online. Now, maybe there's something about how we're formatting it, or getting it up-to-date that's not meeting their needs. So, now, we want to be the leader in using technology to get all this information out, and if we're falling short on that, you know, please let us know. Send mail to Tod Nielsen.

[Laughter.]

Question about automatic feedback to Microsoft for product problems.

MR. GATES: Actually, in some of the preview beta versions, there is exactly that capability. Now, that's a new thing for us. IE 4 was the first time that we did that. But there was even an ability to take -- it would save a lot of the machine state and package that up in a Dr. Watson or some file format so that we could try and do the diagnosis. And we should be doing more of that.

In the case of Windows NT, the file it sends is pretty darned big. So I hope you have a high-speed connection. But for the people that participate in the beta tests, we appreciate the incredible effort that goes into that, and letting you get that feedback up in a simple way. We're certainly interested in making that work better. Once we get the product released, if people are finding a problem, we have a very well-defined process for how the develop group gets pulled in to understand that. How we publish that on the Web, and then how we get it into the service pack. You know, we're very anxious to know for which products you feel like we're doing that well, and which ones you think we need to put even more into it.

Question about what Bill considers the most popular language.

MR. GATES: Well, I think for serious developers, C will continue to be the most popular language for a long, long time.

[Applause.]

MR. GATES: And, you know, we're going to take things like garbage collection and let you use it when you want to use it, and not use it when you don't want to use it. We're going to make the object syntax in COM+ a lot better. So, you know, C is very, very entrenched. Windows NT is written in C, you know, 90 some percent of all the development at Microsoft will continue to be in C. It's at a level that let's us optimize and work with things.

Having said that, numerically we have more customers who use the Visual Basic in Access and the Visual Basic in Office, and the Visual Basic in Visual Basic. Those numbers are ten times greater than any other development base in the entire universe. I mean, Visual Basic certainly wouldn't win the contest of usage in this group. But if you go to the world at large, it is the most used thing. Partly because Access and Office are just so broadly distributed, and we've really coupled those things together.

So, we're going to keep evolving Visual Basic. We're going to let you build objects in Visual Basic. We've connected up the back-end to the same compilation capability. And with the Visual Studio approach, we're sharing more and more between the different languages. So, whether it's Visual InterDev, Java, Visual Basic, or C, all of those things work there. And we're even letting third parties with other languages hook into that design time framework.

And so I think you will see some developers who are mixing different types of code, COM+ is a bridge that allows that to happen. We're not trying to dictate a language. Our message to this crowd will continue to have, C as sort of top of the list, Java, and Visual Basic one notch below that. Our message to the world very broadly, because of the types of applications they do, will continue to have Visual Basic at the top of the list.