Remarks by Bill Gates
Microsoft Corporation
WinHec '98
March 26,1998
Orlando, Florida

MR. GATES: Good morning. It's fantastic to see the participation again this year to discuss the future of the PC. It's been an incredible year for the PC, particularly for customers of PCs. The unit lines continue to go up. We've defined new price points, and there's an increasing number of scenarios--areas like digital photography that people wouldn't have thought of using a PC before that we're becoming better and better at handling.

The advances that all of you provide on the hardware side are really quite incredible. The pace is faster than ever, and across the board, this will make the PC a much better tool. The speed of the processor, the size of the disk storage--these have gotten to the point where you even have some people asking, "Do we need all that capability?" Well, the answer from Microsoft is absolutely yes.We need all the power in order to make the PC a simple device, a natural device to work with.

One of the big areas of advances this year came in flat-panel screens, where the prices came down enough that in some cases, people are even using those in desktop configurations. We need to go much further with this in order to create a flat-panel device that has enough resolution and portability that we can think of one of the key piece and form factors as being a tablet.

In fact, we've been doing a lot of studies on what does it take before the PC becomes a device where you actually do most of your reading off of the PC screen. Today, for some documents--like an encyclopedia--there has been a massive shift away from paper-based reading to a PC screen, and we'll start to see a massive shift in terms of eliminating paper forms and moving them onto the PC screen.

But for more traditional types of print information, we certainly need a lot better user interface and a dramatically better screen. The only question is when that will come, not if. We want to make sure that all the work we're doing enables that.

An area of storage, DVD, is a huge advance, allowing us to have the capacity now to deal with digital video. At first, this will just be plain back digital video, but, as we'll discuss today, we're going to move on to scenarios where the user gets to manipulate the digital video, as well. If there's any area I have a concern for the industry about how quickly it's moving, it's in the area of high-speed connections to the Internet. Although I see very good progress in connecting up businesses at high speeds with reasonable prices, in order to connect up the homes and have a connection--you don't have to wait for it to be there, and it's very, very fast--the only product being made is somewhat of an increase in cable modems and ADSL trials.

But there's a lot of pieces that are going to have to fall into place before we're going to have that kind of availability at prices that will drive it in to a high percentage of homes. And so that's an area of particular focus, where we're working with anyone, whether it's satellite technology companies or phone companies or cable companies, to help drive that forward.

Part of the vision here that will make the PC a great tool in every house is a concept we call Web lifestyle--that is, thinking of the Web as a mainstream way of communicating, the place you turn when you want to find out the latest news, sports scores, when you want to buy a gift, when you want to learn something, whether it's about a medicine or a hobby. The Web should be the obvious place to go. If you want to organize a trip or get together people who want to participate because they share a common political belief or interest, if you want to stay in touch, send electronic mail--even new forms of entertainment, multi-player games across the Internet--all of these things, it should be common sense that you turn to the Web.

But in order to do that, we need even more rich material on the Internet. Now, the levels of investment there are pretty incredible, particularly given that the business models for those investments won't be very strong until we have at least an order of magnitude in increase in usage. That will start to drive mainstream advertising and purchasing volumes and will make these investments in Web sites pay off in a worthwhile way.

This is definitely coming. Today the only population you could say that lives a Web lifestyle is probably students on university campuses. I also may fit that category, given that I've got a 45-megabit connection to my house and reasonable flat-panel displays in every room of my house, and I have to say, at least based on my experience, the Web lifestyle is a fun thing, something that we should try to make affordable and reasonable for everyone.

Now, the PC will come in many different form factors as this rolls out. The primary form factor will be the PC we think of today--the full-screen device that runs all the applications. Whether it becomes a tablet-sized device or is a CRT-type device that will always be the mainstream, the place you go to edit documents, the place you go to fill in information.

But when you want to just browse information, smaller form factors will make sense, particularly as we make your information more verbal. Now, that's a big initiative we've undertaken so that you don't have to think about, when you have multiple PCs, whether your files are on one or the other or your favorite preferences are on one or the other. We're going to make it so our laptop flows, because of the operating system, through any connection--infrared, the Internet--so the information that you update is available through all their machines.

Because of that, these small form factors make even more sense. The hand-held PC came out about a year ago, and we've got about a half-million of those in users' hands. The real runaway best seller has been the new color screen model--I'm showing the one here from HP, but also Sharp and many others have come out with that kind of device. We just shipped version 2.0 of the hand-held PC, with a lot of additional features based on user feedback.

A new form factor for us comes in a few months. That's the Palm PC. This is where we actually use handwriting and the touch screen for input, and we expect that that will also be very popular.

The Auto PC was introduced in prototype form at the Consumer Electronics Show early this year. It'll be out as an add-on for cars in the United States sometime this year, and eventually car manufacturers will build it in as factory-installed equipment.

It's been amazing the reaction we've had to this.We thought of the Auto PC as kind of a specialized form factor and one that would take a while to capture people's imaginations, but the idea of using speech recognition, having the traffic data, being able to replicate your name and address list so that addresses that you might want to get instructions how to travel to, they're just there, and you can collect them--all of that has people pretty fascinated. People will be able to actually make use of it as we get that out to the market.

A final form factor is the intelligent TV, and there's no doubt that this will be using PC technology--PC processors, PC graphics and video. The only thing that is not necessary, of course, is a new display or a local storage device when you've got the high-speed, two-way connection.

Web TV is a very great start there, because we've brought some of the ease of use and benefits of integration to that TV-watching experience, particularly with the Web TV Plus that we brought out last fall. You'll be able to think of Web TV as simply a very low-end PC. It doesn't run all the applications, but it gives you the Internet access, and then at retail you'll have a family of products going out to the very high-end PC itself.

One of the big challenges is making sure that all of these things work very, very well together, and the industry, particularly Microsoft, has a lot still to do here. We've put together a video that captures what it's like today to put these devices together, so let's take a look at that.

[Videotape spoofs ease of use of TV and Web devices together.]

MR. GATES: Well, we've still got some work to do.

One of the challenges we have with PCs is that there's a lot of concepts that people have to deal with--concepts of the name space in browsing the Web, the name space in browsing of their mail folders. The file systems and different search commands that it has, versus all those other things, and by bringing these together and having one way of browsing information, we can make the system far simpler.

A good example of this is that in Windows 98 and Windows NT 5, we take the help utility, which had its own special command, it's own proprietary file format, and we simply get rid of it. We leverage the browser capabilities that are built in the system, and use that for all the help capabilities.

Now, in order to make that work, we've had to enhance the browser itself, we had to add some things to the HTML capabilities so we'd be a superset of what helped us out in the past--the Window management, the command simulation and driving, but we simply eliminated another thing to be loaded in the memory and another set of commands that the user would have to learn.

Now, we'll go even further in this unification. We will get down to a single shell that can take you everywhere.

Another big problem has been upgrading systems. It's been risky getting the latest drivers, knowing that things are fixed, and so now we've built into the operating system the concept of going out to the Internet and getting the latest updates in an automatic way.

Another pretty straightforward thing we've done is take a look at all the different error messages we have in the system, and we're finding that even we don't understand all of those messages, and exactly what the user's supposed to do.

This is one of my favorites, and, you know, I actually know what it means, but it's still kind of strange: "The DHCP client could not obtain a drive. If you want to see DHCP messages in the future, choose Yes. Otherwise, choose no."

Well, "otherwise choose no"--that's so helpful. I mean, this means, "Hey, I love seeing these messages. I know exactly what to do when I see this message," and so instead of thinking of a message like that coming up, we should bring up something that starts to diagnose: where is the connection broken--and really guides the user to do things, and so the error text and the help messages should all be unified, and, in many cases, we should be able to get rid of those.

And so there's a lot that needs to be done, quite a bit of it at the software level.

Now, one of the scenarios we tackled in Windows 98 is digital television. We include in the programming guides and the ability to download that by using a broadcast TV schedule. Here there are some really neat things that can be done, and it's a step towards what we call digital television.

"Digital television" refers to not only having higher quality video, but also having the interactivity, having to access more information.

So let's go ahead and take a look at this. The broadcasting format that we're starting with is called HD 0, and that includes not only today's interleaf formats, but a high-resolution progressive format, either 480 P60 or 720 P24, and that lets you do movies at very high quality, and it's quite impressive what that looks like.

This is what we want to have built into every PC, and we've been working with the TV equipment industry and the broadcasters to explain why, no matter what format the production work is done in, actually transmitting in progressive is a very good approach, because it's the most efficient way of using the spectrum, and that means whether it's for multiple channels or the data enhancement, you've got extra capability there that can be used.

So let me ask Steve Guggenheimer, who's the product manager helping drive these capabilities, to come on out and show us what digital TV looks like.

MR. GUGGENHEIMER: Come on over. Have a seat--a chance to relax.

As you mentioned, digital television has about three key features, and we think that in order to make digital television a reality in the near term, we have to do two things. We have to enable a high volume of reasonably priced receivers so that consumers go out and buy the receiving capability, and we need to offer some form of business model or business opportunity for the entire television industry-- broadcasters, cable, et cetera--so they can recover the cost of moving to a visual infrastructure.

So on the PC side, we're all very familiar with information access, so I brought along the other two things. It does the high-definition or the HD0 capability, as well as interactive programming.

Now, this particular PC has the hardware in it right now, and is receiving, decoding, and displaying a 480 P or an HD0 level signal, and it's the PC that's now outputting into this TV as well as the normal screen.

The hardware for this is something that's actually going to come up in the next year. In fact, ATI has said that we can announce this morning that within the next six months, they plan to try to put on their mainstream graphics chip the ability to receive and decode the HD0 capability so that mainstream PCs within the PC '99 timeframe will have the hardware capable of supporting this level of video so consumers can get that.

In terms of the software, I'm running this particular machine on NT 4.0. We're running the same thing next door on Windows 98, and we're even adding capability into Windows CE so that we can offer consumers a range of receivers, going all the way from the set-top box form factor all the way up to the work station form factor. So in terms of enabling a high volume of receivers, this type of capability will be available on a large number of receivers of all form factors within the next year.

Now, let's talk a little bit about interactivity. You'll notice in the lower left-hand corner, I have an I. So as a consumer, if I wanted, I can simply ignore it, or, if I'm interested, I can go ahead and bring up the tool box.

So now what I can do is I can take the broad reach medium of television and combine it with direct interactivity of the Internet. So now as a consumer, if I want statistics, I can simply click on statistics, and they come up when I want them, versus the broadcaster sending them to me before every commercial.

Going back to the toolbar, another interesting model is replay. So as a consumer now, if I want a replay, I can get the replay once, I can watch it over and over again, I can come back to it later in the game. And if you think about it, that requires storage, another common component of the PCs that we're all familiar with.

Now, in both of these areas, you'll notice that there was advertising associated with this, so in terms of a business opportunity, there's the opportunity for more or new forms of advertising, and that's very familiar to broadcasters.

However, in addition, there's other things. For example, transactions. In this case, I could buy a jersey associated with this game. Or, if you were watching a television ad, you could be watching an ad for CD-ROM and go all the way from watching it to purchasing it. So it gives both the broadcasters and the advertisers and the community the opportunity to move from just a push media all to a full medium or to transactions.

The last on I want to show you in terms of being a little bit creative. You talk a little bit about potentially having multi-channels. Well, what about a subscription program that says the television program, the football game is free, but for an extra 50 cents a month, I can get a second camera angle. This has enough bandwidth to support two complete pictures. I could get the one camera angle for free and then for an extra 50 cents get a second camera angle, or I could get the local commentator instead of the national commentator.

So what I wanted to show and what we've done is basically that we will be enabling digital TV capabilities on a broad range of receivers, ranging from the PCs down to set-top boxes over the next year and, as you can see, it's quite a nice thing, and I'll go ahead and just hit B to demonstrate. It's on the PC.

MR. GATES: What we saw there was watching high-definition video and being able to interact with it. Another scenario that we think is very important is actually letting people create their own high-quality video--to be able to do editing right there on the PC. So in the same way that we see still photos and keeping albums of those, mailing them around--we see that as being popular--we'd also like to bring in a motion video.

Well, there's a very tricky element here, which is that motion video requires very high data rates and in order to deal with it at all, you have to have real-time compression, real-time encoding into one of these compressed formats, and that hasn't been possible in the past. In fact, it's only been with very, very expensive workstations that people have been able to do that.

Now, through the miracle of technology, we can see that that can become a future of the PC, so real-time digital recording. I'd ask Peter Biddle to come out and show us how this is going to work, how we're going to get video in as a first-class data type.

MR. BIDDLE: Thanks a lot.

Okay. What I have here for a demo is I have a Toshiba laptop and it's in a docking station, and inside the docking station is this new chip from C-CUBE--it's an MPEG 2 encoder/decoder chip, and plugged into that chip, we have an actual video feed from one of the cameras back there. We also have a Pioneer DVR drive connected to this system.

And what I actually just did is when you were over there on the couch, I took a feed from that video camera, and we should be able to see this come up on screen. So what you're seeing on the screen is MPEG 2 data that we real-time encoded using this chip onto the hard drive without playing it back, and this happens real-time.

MR. GATES: We're even adding capability in the Windows CE.

MR. BIDDLE: In the audio.

MR. GATES: Yes.

MR. BIDDLE: Okay. So not only can this chip encode, it can also decode, and it can decode two simultaneous streams. That has some sort of immediate ramification for end users, that the first one we want to demo is playback with two streams at the same time. And here we go.

So what we're seeing right here is that video once again, but I'm going to add "Michael Collins" from Warner Brothers. We're going to stick that in the mix. And, as I said, because we're doing two streams at the same time on the system, we can start doing per pixel transitions between the two, so you're going to see here is a set of programmable phase wipes and other transitions that are only capable if you're capable of doing two streams at the exact same time. We're not switching sources, we're actually decoding both streams simultaneously.

And this also has interesting sort of repercussion with picture-in-picture, where HDTV has, you know, very immediate benefits there, as does single-click recoding in an electronic programming guide on like a Web TV or a broadcast PC. You could go away for the weekend and stack up all your shows and digitally record them.

A system with this level of functionality would have cost about $100,000 a year ago, probably well over that, and this is going to debut at consumer price points, but what's really interesting is that we also have a DVR drive on the system.

So just to review, we took the signal, we MPEG 2 video encoded it, we MPEG 2 stereo audio encoded it, we put it down on the hard drive. Then I ran a program that turned it into DVD video format, and then we copied it down. So I'm going to pop the DVR into the system here.

So you can see this disk right here, here it is. And, as I said, we went from end to end here, and let's take a look at what happens. This is a consumer DVD video player, so we take the disk, we put it into the consumer DVD video player, and here's an act of faith. So what did we just do? In four minutes, we went from nothing, analog capture, MPEG 2, onto the hard drive, turned it into DVD video and burned it on a DVDR disk, put in a DVD video player and played it back.

MR. GATES: I'm certainly surprised to see that C-CUBE is getting that chip out and getting it out at very low price points. So this is one capability that's going to actually appear faster than we would have expected.

Well, for Microsoft, this is a very big year. In fact, this is the first year where we've had two major activities in the Windows product line. By mid-year, getting Windows 98 out and also, about the same time, getting the second major beta of NT 5.0.

And those products represent a massive amount of work, many of which are realizing initiatives driven by this Windows tech forum. For example, getting in the new device-driver architecture, the WDM architecture. Windows 98 and NT5 implement that architecture, and that's going to allow people to do things with real-time audio and video and not have to kludge around the operating system when they want to get those capabilities. We're very exciting about getting that foundation in.

That's also the place that we get into to rich manageability capabilities built into the operating system, so that's a big step forward for all of us.

Other key products from us are Microsoft Office and Back Office. In fact, if you take Windows Office and Back Office, that's the overwhelming percentage of everything Microsoft does. And because we have so much opportunity to integrate more capabilities into those three products, those will remain our primary products for a long, long time to come--certainly all the way out to any time frame that we can make predictions for.

We're increasing the R&D we've got on those products pretty dramatically, as well as bringing in some new interactive media activities, where we're pioneering things out on the Web, like many other companies.

Perhaps the most interesting thing is what we're doing in initiatives to span all of the products. In 1996, it was very clear to the world what we were doing. We were building Internet capability into our products very broadly. That was our top priority, and we've come a long way on that initiative. Whether it's the popularity of our browser and how that's done and all of the reviews, the popularity of our Web server and the way we've built that into Windows NT--all the tools like visual inter-gab or Front Page that are by far the most popular within their categories.

There's a lot more we're still doing on the Internet, but we've got that so well understood that we don't have to have it as our top priority, just one of the top four things we're doing.

On the Internet, security is still a big issue. Moving up to ATM support and quality of service capabilities, moving to ITV6. A lot can be done, which is why you find Microsoft so active in all the Internet standards committees. NT5 is a big milestone for many of those things, with its ATM and QOS [quality of service] support.

Another exciting thing along those liens is the work we've done with Cisco on Active Directory, to make sure that if their communications equipment needs to verify who a user is or what the policies are, they can link in to our active directories. And out of that come some industry standards that all communications equipment can connect into.

In 1997, our top priority became the focus on simplicity, part of which is cost of ownership. About half of everything we've done in NT5 comes out of that key focus. It's something that I think will stay a top priority for a long time to come, because there are many, many aspects to it.

Two other key parties are the scalability and manageability. Now, scalability--we have come a long ways. The days when people thought of the PC as a low-end device, those days are gone. In fact, if you look at sales last year of UNIX workstations versus PC workstations, you can see a huge divergence. UNIX workstation sales were down last year, whereas PC workstation sales were up over 80 percent.

We just had a great forum that Intel and Microsoft sponsored, which was the workstation forum, and there we had a chance to bring up all the people who do high-end software in the niches where performance has been so critical--people who do real-time trading software, mechanical design, electronic design, all of those people coming in and showing off the work they're doing around Windows NT and Intel architecture. And with those developers, the opportunity to sell in volume is very attractive, and now that they see the performance there, they are moving very quickly. In fact, that's another case I'd say that's happening faster than I would have expected.

At the server level, there's still quite a bit to be done. We got our clustering support out last year. That gives people a two-node fail over capability, but we need to crank that up to allow even broader clustering capability for very high performance.

There's an approach there called "shared nothing" that we think will actually be the most popular way that we'll be able to drive up -- up to what the most demanding applications have required and well beyond it, because applications on the Internet are going to expect even more.

And so a combination of higher chip speeds, more chips per server, and clustering those things together, along with software improvements, will lead to dramatic gains, and not only will the PC not be viewed as the low end, it will come to be viewed as the high end, and that's where the focus of development for all new applications will come.

Now, NT5 is a big milestone for us in every one of the initiatives I talked about. One of the breakthroughs here is in this idea of state management. Your state as a user has been very difficult to migrate. If you think to yourself, "Geez, I'd like to move from one PC to another," you have to think well, where is my list of favorite websites? Where is my list of spelling dictionary words? Where are all the files that I created?" And getting that over to the other PC would require a lot of work, even an understanding of such mysterious things as the registry and everything that goes on there.

Well, we decided with NT5 that state migration had to happen just as a feature of the operating system. It would lead to a lot of benefits -- being able to roam, being able to go home and use your home PC and get it all up to see -- and eliminating the need to ever have to come to your individual PC to administer that state. It would all be stored centrally and subject to whatever policies that administrator wants to set.

And there's a way to do this that would have been a step backwards, and that is to simply move everything back to the center. If you do that, then you give up everything that's been good about the PC -- the responsiveness, the peripherals, the portability, and, in fact, the approach that was proposed by the -- let's call them the anti-PC crowd -- would have required rewriting all the applications, and that was the so-called NC movement, or, as we like to say, the non-compatible movement.

Well, moving the state to the center, that would be a huge tradeoff against things that PC users expect, and so our approach, which we call IntelliMirror, gives you the best of both worlds. It gives you the state down on the local disk, so that it's always there for performance, but it also logically stores it in the center, and so we give up nothing to get those benefits of working centrally.

So this is -- has really gotten people's attention in TCO reduction. The Gartner Group now has shown how people even with a zero admin kit can get significant benefits, and we're very pleased with the work that the hardware industry has done with Net PC and the Windows-based terminals. The Net PC is our attempt to get rid of some of the drawbacks of the ISA box and have all the devices be auto-configured and identifying.

Windows-based terminals, for people who are very light users, where you actually not only move the state but also the execution up to the center without giving up the user interface or asking things to be developed in a new way.

So let's go ahead and take a look at NT5 IntelliMirror, and I've asked Valerie See to show us one example of where the IntelliMirror can help out pretty dramatically.

MS. SEE: Good morning. What we're going to show is what happens when a machine fails, and we're going to do a side-by-side machine replacement.

So the first thing I'm going to do is -- Bill's logged in on a Compaq Net PC, and we're going to do some real-time changes to the desktop. I'm going to go ahead, pick a wallpaper. We'll move some icons around, just to show that even when something bad happens, none of the local user state is lost, it's all being mirrored to a central server, as well as being preserved in the local cache on the hard disk.

So let's say, you know, you log off for the night and, you know, something bad happens to your machine -- you know, a meteor hits, or something like that. We're going to have something bad happen to your machine. I just pulled the power out. Well, all right, so you come into your office the next day and you say, "Geez, this is bad. My machine crashed." You get on the phone and say, "I need a new machine."

So let's plug in a new machine. This is another Compaq Net PC. This one's a little better. This one has actually got audio. So what do we do to get this guy going?

MR. GATES: Now, this machine -- what is on the hard disk here?

MS. SEE: Absolutely nothing. What the IS department can do is just wipe it clean, bring it to his office, and the first thing we're going to do is turn it on, and the only thing that this machine is going to be booting from is the network, because it's all that it's got to boot from. There's nothing in the hard drive, there's no floppy drive in this machine.

So what you're going to see coming up on the screen after the BIOS boot message has cleared is you're going to see the network boot prompt, and you're also going to see some messages about where it's going to go out, find a DHCP address, find a boot server, and then start pulling down code that's going to facilitate the replacement of the operating system, as well as your user state.

Now, because there's some secured access going on here, as we drop this new machine into the network, the first thing that it's going to need to do is we're going to say, "Okay, this guy doesn't have a machine account. We've never seen this machine on the network," so we'll have to log in as a user with sufficient permission to say, "I want the machine account that was assigned to this guy to now belong to this guy." So that's going to be the first step in this process. After that, we'll follow it up with, you know, transferring user state, formatting the hard drive, and pulling the operating system over.

So this is the OS IntelliMirror introduction screen, just to kind of let you know what you're going to be doing, and this is the point at which we need to log in as a user with permission enough to take this guy's machine account and transfer the rights to this machine. At this point, we get a menu of options, and what we clearly want to do is replace a failed computer. This one's had a meteor strike.

So now we need to know what machine account this guy had, and his name was Compaq 4. And so we'll make this guy Compaq 4, ask this to just confirm that, make sure that we don't nuke anything, and at this point, we've transferred the machine account over. The next thing I'll do is go ahead, transfer the user state, and get the hard disk formatting.

If you want to go and look at another demo, by the time you get back, we'll be ready to log in again.

MR. GATES: Another area that we made big advances in is in networking infrastructure. We've got the telephony network, we've got the data networks, you've got video networks -- all of those have been very separate, and to bridge that, you have to take standards from the phone world and the IT world and try and hide those from applications.

The goal is very straightforward. It's to let people right rich communication applications on the PC that can drive all the capabilities of these machine networks. And TAPI 3 is a big part of that. We've had TAPI for a long time, but TAPI 3 is a dramatic expansion in its capabilities, and we're doing this in a way that it works with all the different networks, all the different transports.

And the best way to understand this is to see some examples of it actually being used, so let's go ahead and ask Petra Nicodam to come on up and show us TAPI and what it enables.

Good morning.

MR. NICODAM: Good morning.

What I wanted to show you is some of the new networking features we have in Windows so that you can use Windows as a platform to fill a network for rich communications.

Before we actually get into the demonstration, I'd like to walk you through the configuration we have set up as a demo. If you'll look at this diagram here on the middle screen, you can see that on the left, we have a small corporate network with two PCs connected to it. They are the machines I'm using and also a few of the machines here on my right.

One of the PCs on the corporate LAN is a server with two network cards in it that connects the server to the corporate LAN and an ATM card that connects the server to an ATM switch, which in turn connects the corporate LAN to a public network. There is also a telco central office server. The central office, in addition to traditional telephone service, provides fast Internet access by ADSL service to its customers. There is also a small office or home with a PC and a telephone. You can see the configuration on the front left corner on the stage.

The phones, the PC and the phones are connected to a public network over the same physical phone line, and the ADSL deployment we used here in this demonstration is based on PPC over ATM over ADSL architecture, and the equipment we use for both the central office and the customer premises are Alcatel.

Let's pretend I own an advertising agency, and I'm preparing a video clip for my customers for them to use it in their commercials.

I know that several of my employees, they are working on this video and discussing the video in the multipoint conference that is happening in the corporate LAN. You can actually see that I have the two people there. I can see and hear them. They're participating in the conference.

[Mr. Nicodam discusses the status of the video with the two people and plays it from a video server while still conferencing with the others. He then uses the same application to play a voice call from his PC over a voice modem to a regular phone that is at the customer's premises. He talks to the customer and has him log in to play the video remotely over ADSL.]

MR. GATES: Now I'm going to quickly check in with Valerie. I think we gave her plenty of time to get my machine up. How's it look?

MS. SEE: It's been waiting for you for about five minutes, so it's ready. Let's go ahead and log you back in and make sure that the state changes that we made to your user state, the wallpaper change, the icon movement, it's going to go ahead and still be there.

Additionally, this new machine should have sound and, hopefully, we're going to hear some log-in chimes, which we just heard as you logged in. So you got a better machine, even though the meteor struck your old one.

We notice that the wallpaper is, in fact, still there. The machine is going to be plug-and-play and your monitor, your new audio, is going to be just, you know, finishing up. While this is all going on, you can be doing work, but the cache on this hard disk, which was empty, is being filled in the background, and you don't have to wait around for that to bring all of your data over. You can still do work and go ahead and get things done while the cache is filling.

MR. GATES: I think the big breakthrough in PC computing will come through a natural interface, and we're increasing our investment in this area quite dramatically. We added Cambridge, England, as a research site. We'll be adding others. And we've ramped up very quickly and are making excellent progress in these key areas. The way we describe it is pretty straightforward--computers that see, listen, and learn.

It's more than that, of course. We've got to have much better graphics, we've got to have a much better way of doing software development, a much better scaling, but the things that will really open up the market are the language understanding, speech understanding, and handwriting recognition.

Each of those, it's fair to say, are--we are able to take early implementations and ship them already. Our linguistics support that's in the word processor and in the new grammar-type capabilities and a big advance in that was the version of Office that'll ship this year.

Our speech technology, some of which we do in collaboration with Lernout (ph) and Hauspie (ph), that will be in the auto PC. The chip in the next major release of Office will also be directly exploiting quite a bit of that.

And so already we're able to get feedback and use these things as the hardware comes up to speed and we get these approaches built in with rich APIs.

I just wanted to quickly show a piece of this. The thing that I think is hardest for people to think about is this idea of learning, the idea that a computer, after you've used it for a year, would actually be easier to work with than the first day you pulled it together.

This is actually a bit of an agent that's parsing natural language, and because you're doing your interaction with mail messages, it can chime up and help you. So I've asked Phil Fawcett to come and show us what learning agents might do for us.

MR. FAWCETT: In Microsoft research, we've been spending time with users learning what language they use when they want to schedule events and what they might--how they might interact with each other in order to make a point.

The technology we're using here is a number of things integrated. We're using text to speech, we're using voice recognition, we're using natural language processing, and we're using the ability to use an agent and get a context out of the message.

So it's a fairly complex situation, so let's give it a go.

Here on the screen is just an example of someone sending me an e-mail responding to a schedule request saying, "Hey, I can't make it, but how about a week or two later?" Now, what you'll see here is an agent coming up that asks me would I like to see the calendar. I'll say yes.

So you can see, Bill, it picks up the context of a week or two later, it brought up my schedule. Now it can make an appointment.

Now, as you're aware right now, making a Schedule Plus, you know, takes 10, maybe eight steps. But in this case, my hands were free, we were able to be--have a conversation going on. I said yes. It's very efficient.

So let's try another example. And Eric's asking me to go for a board meeting.

MR. GATES: So the agent was reading that.

COMPUTER: Would you like to schedule an appointment?

MR. FAWCETT: Yes.

COMPUTER: How about Saturday at 4:30 at the Inn?

MR. FAWCETT: Now, there's a number of things that happened here. It read the message, it read the fact that I want to do a conference call for the board meeting on Saturday, and notice I used an abbreviation. I didn't use the full word. It picked that up. It also gave a context. So now it read the fact that we wanted to do it at 4:30 Pacific Standard Time, and it brought that up and set an hour approximately for us to have that meeting.

Let's go through one more example. And normally I'd say yes, and it would actually show up on my schedule plot.

This is an example of it picking up the word "June," the word "schedule," and bringing up my entire schedule for the month.

COMPUTER: Would you like to see your calendar?

MR. FAWCETT: Yes.

COMPUTER: Here's your month of June.

MR. FAWCETT: Boy, it's a busy month. All work.

So you notice that it brought up the whole month, and with schedule information I could go through and make the appropriate appointment.

And, again, this is an example, Bill, of how we're applying research to make the user experience better and also to move the industry and user experience in a more quality way.

MR. GATES: Great. Thank you. That's super.

Well, WinHec has been a great success in the past in driving forward key hardware initiatives--the hardware guides, the plug-and-play, UST--all of these, we're really seeing the payoff from joint industry activity.

Moving forward, things like the 3-D graphics display in our direct-text APIs, we feel like there's been excellent progress there. Home networking--I'd say we're at the beginning of that, because there are many proposed approaches, but I think that's a critical thing to solve, to really push forward consumer PC use.

And finally in the area of communications, all these telephony standards that we need to get drivers for and make sure that users don't have to understand what's going on underneath.

There's a lot of work to be done working together to make sure the quality of hardware drivers is improved, as well.

So we see a lot of opportunities, looking forward. The Web lifestyle is going to require a lot richer PC devices, as well as the new form factors. We're going to need a lot more performance in order to enable video editing, as well as of a natural interface, and we're really pleased with the partnership approach that the PC industry takes. It's quite unique to have the intense competition and yet working together to improve the fundamental standards and really grow the business.

This is something that we feel honored to be able to sponsor and bring together so many innovators, and we're also very interested to have your feedback on how we can continue to do this and keep improving it.

It's an exciting time, and we thank you for coming.

Thank you. (Applause.)

MODERATOR: Thank you very much, Bill.

As is our tradition at WinHec, we've collected the questions from the attendees at WinHec in the question box since Tuesday evening. Because we're running a little bit late, I think we're going to limit ourselves to the questions on the cards and not be able to take questions from the floor mikes, because I think you have another appointment to go to.

So let's go ahead and get started. When will we see speech recognition in Windows operating system products?

MR. GATES: Well, the speech recognition will start off like most features do, as an add-on product, and you're seeing a couple of those out there today. Microsoft will be doing an add-on product. It is built in for the Auto PC, and we are working on a set of APIs.

It's a particularly interesting API challenge, because, with a probablistic input system, you need essentially a description of what's likely to happen, coming from the applications, being matched up with the input system and the probability system.

We're continuing to refine that, and I think by next year, we'll have a very rich API and then in the next major round of Windows, will have it as an optional but not required capability, and then maybe the major round after that will simply rely on it as the primary interface.

MODERATOR: We have several questions that we got about the strategy for Windows 95 going forward and how it relates to Windows NT and Windows CE. I'll pick one formulation here. "As I know, Windows 95 will be merged into Windows NT, so is the Microsoft operating family Windows NT and Windows CE? Is that right? Or, if not, please describe it for me."

MR. GATES: Well, we're very enthused about Windows 98. In fact, Windows 98 is probably the simplest upgrade that we've ever provided, in terms of not requiring more system resources, and bringing a lot of key benefits based on what people have asked for over the last several years.

But as we do the next major round, we are going to start with the NT technology, and so we'll need to build a form of NT that's aiming at the consumer market. We're still working out exactly what that will mean in terms of the features set there, but having one kernel, the NT kernel, be at the center of all our activities. It's pretty important to us.

Windows CE we see on the new form factors, and in the past those form factors, you haven't been writing applications for them or they've been operating systems that have no relationship to Windows. We think by having the same kind of API, the same kind of driver structure, we can have some sharing there. And so for the next several years, Windows has three tiers: Windows CE, Windows 98, and Windows NT5.

Now, Microsoft is sending a pretty strong message to business customers that if they've got the memory to run NT5, they're better off to plan for that, and we do think we'll see the market on the business side shift to NT5. Even some of the high-end consumer could shift to NT5, but it's the next major revision that will get down to a flavor of NT aimed at consumers and a flavor of NT aimed at business.

MODERATOR: Is Microsoft's ability to design and incorporate new technology into future operating systems really at stake with the Justice Department, or is the issue about maintaining a fair, competitive market for everyone?

MR. GATES: Well, it's an interesting situation. Basically, what they're asking us to do is, in addition to shipping our full-feature products, which have all the capabilities, they're asking us to ship a crippled version of the products, where we actually either hide or take out some of the capabilities.

And if you think of every new feature we add to the operating system, we add well over a hundred different things, and certainly in every case, there's been some add-on, somebody who was doing something in that area as a payoff, they got to say, "Hey, I'd prefer not to have that in the operating system," that that would be nice for them.

If you think of the number of versions we'd have to ship and how difficult it would be to test those or have a message to developers if we had to pull those things out, that would be a big step backwards, and so although where it's the browser, you can go ahead and do the crippled version that doesn't pull you down, the general principle of not being able to do the integration and depend on that integration, that would be a disaster for Windows. It wouldn't allow us to do what customers are asking us to do or developers are asking us to do.

And the browser is a perfect example of this, because the simplicity it brings, what it lets you do the applications, those things are very important. And if we're going to make sure we go out and really articulate why we're doing that for the developers and what it means for users and make sure that we -- it's clear the very, very broad support people have for us moving in that direction, with that, we're confident that this ability to innovate will be preserved. Certainly in every case in the past where this has come up, the law's been very, very clear on that.

And so our statement here is that nobody should be concerned about this, although a critical principle is at stake.

MODERATOR: Is Java a mainstream strategy for Microsoft, or are you just hedging your bets?

MR. GATES: Now people finally understand what we've been saying about Java for quite some time, and that is you can think of Java language and Java's run-time platform as separate things. You can use the language and actually take advantage of anyone's software, or you can write in the least common denominator approach, and it just calls that Java run-time.

Anybody who's got software, whether it's IBM or Novell or Oracle, likes people to write applications that calls that software and takes advantage of it. In the case of Windows, we think people like applications that are efficient, because they graphics APIs, they use our user interface, use our clipboard, and therefore it is the Windows environment. And I doubt any application that doesn't exploit Windows will ever sell well in that environment. It's up to the marketplace to decide that.

The tools we put out, our Visual J++, give people a choice. With the Windows foundation classes, you can call directly into Windows and get the user interface and the performance there, or you can continue to call just the Java run-time. That's an option you've had in every language that's ever come out -- COBOL or C -- you could always restrict yourself and make the tradeoff to get more portability.

For commercial applications, I think the proof has been pretty strong. You've had people like Corel and Netscape try to take the least common denominator approach and, you know, I think the beta software's still out there if you want to remind yourself what that was like for those applications.

So Java's got an important role. We work hard on that. If you look at PC Magazine for a view, you know, of who's doing the best job in terms of speed and compatibility, it's definitely us. The fact that we want to give people the flexibility with Java, let them do whichever thing they want, certainly Sun doesn't like that, but we're giving people a choice of languages and when you use the languages, a choice of whether you call and exploit or whether you stay just to the common run-time.

MODERATOR: Bill, does your company let you do any software programming anymore?

MR. GATES: Well, I don't think there's any problem with them letting me do it. Certainly I have plenty of authority to do that if I choose. Unfortunately, I don't get a lot of time to do development. I did play around with the key programming tools, the J++, the latest Visual Basic, and to write some fun applications. I wrote what I call a Web kiosk application, where I bring up four Web pages at a time and I cycle through popular Web pages, and so at my desktop there, I've got the trade magazines and the general news and some of the Microsoft sites and competitive sites cycling through, and that was a fun little application that actually showed me that we need to make it easier to call some of the system's services.

So mostly what I'm doing is sitting down with develop groups and talking about what features we put in, what basic approaches we take, and so there are legions of really incredible programmers who get to write all the code.

MODERATOR: In the technology race, is Microsoft driving too fast for end users to be able to follow? Will Microsoft or the PC industry lose customers for this reason?

MR. GATES: Well, I don't think -- you can't ever go too fast, because the marketplace is there, and it's like are the new things you're doing exciting? The improvements we're making are driven by what people are asking for. I mean, the simplifications and some of these new scenarios. You know, some of these new scenarios like dealing with digital photos in a great way and making the PC the place you want to work on them -- we have a lot of work to do to make that broadly attractive, but the interest level is very, very strong.

Sometimes we get people saying we're moving too slowly, sometimes we get people saying we're moving too fast. As long as that's about equal, maybe we're striking the right balance.

We did really take our time since the shipment of Windows 95 not to be revving Windows again and again and again, to gather all those things together and do a really major release, which is Windows 98.

Likewise, with NT5, we didn't just come out with a version that has a directory and a version that has an IntelliMirror and a version that has the networking. We brought that together into a major release.

You'll see us in the future be more quick about very minor updates that are just sort of packs that you can transparently get across the Internet without thinking about it, and then these releases where you have to think about them, but the payoff is very, very strong, and we think having, oh, you know, a two- to three-year cycle between the major releases, that that's ideal in terms of both enabling the new scenarios and getting customers the time to plan for those things.

MODERATOR: Describe how we'll be using devices in our homes in the year 2000? What features will be part of the PC that aren't there today?

MR. GATES: Well, the year 2000's pretty near, you know. The scenario I'm most excited about is that tablet scenario, where you actually can do your reading and your annotation off a device that you can carry around with you, and it's got better resolution than today's flat screens. That's probably more like five years out in the future.

The year 2000, I hope this DSL, this low-cost DSL that's getting popular, I certainly hope that photography, music downloads, video editing, that we've been able to make those very mainstream. I hope we've made it extremely simple that when you leave your PC at work if you want to get your files or any of the preferences you've set up, that as soon as you get to your home account and log in, boom, that information is all there for you, and so you don't actually have to bring the hardware home in order to have that kind of accessibility.

Certainly 3-D will probably be the biggest change. We are right on the verge of begin able to assume 3-D enough that we can build that into the standard interface. So the next major rev after NT5, I've told people, "Hey, let's be dependent on the performance and capabilities and take advantage of 3-D there." So you ought to be on the verge of seeing those capabilities.

MODERATOR: Okay. I think we've got two last questions that we've got time for.

How do you see the migration to IH64 appearing?

MR. GATES: Well, Microsoft is very enthused about IH64. I remember the challenges of the 20-bit address base and then the 24-bit address base and then the 32-bit address base -- not only the challenges, but the complexity that was involved in mixing and matching applications written to those various things.

It's pretty impressive what you can do with 64 bits of address space, and I think I can say confidently that in my lifetime we won't have too much problem with running out of 64-bit address space.

IH64 is a really impressive project, and we've got pretty deep involvement there to make sure that our software does a great job of taking advantage of that, but even more to make sure that we give people the tools to start doing the sorting. We've already got some things out there. We've gotten good feedback on that, and so Intel and Microsoft will get the developers going by the end of this year, and so when the chip ships, you'll have NT there, and you'll have quite a few of the applications precompiled.

The performance capabilities you get as you bring in more RAM, go beyond 4 gigabytes on the server or some of these high-end workstation things, I think people are going to be quite impressed with that, and so we're going full speed ahead. You'll see everything from Microsoft taking advantage of IH64, and it's a pretty smooth transition. In some ways, it's simpler than going from 8086 to 286 to 286 to 386. This one, you know, we've learned a lot about those transitions and we're planning well in advance much more than any of the others.

MODERATOR: And, in fact, we have a presentation on it tomorrow and we're giving people some preliminary tools to begin some work with IH64.

Final question -- I know you've touched on this -- what's the future of broad-band PC communications and what price point or cost will households be expected to pay for high-speed access?

MODERATOR: What's interesting is the price of the PC has come down so much that your monthly Internet connection fees are actually very significant. You know, paying $20 or more a month, actually the present value of that is much higher than the cost of a very powerful PC.

And so we're definitely going to run into some price sensitivity there. The ideal would be to get well under $20 and have the kind of speeds we're talking about DSL and the kind of immediate access so you don't have to wait for the connection and tie up your phone lines.

I think we're going to take a big step forward with that. Having the DSL available, being able to use the DSP that means that you use the same hardware to do analog modems, home networking, and the DSL means that the hardware cost is not very high and we've gotten Compaq to take a leadership role there and drive that into a number of their PCs.

So although that will start in '98, by '99, I think the pieces will be in place there that the phone company will start to see it's a high volume opportunity. They're being held back a little bit by the regulatory environment, and so I think the industry has to provide them some support to make sure that DSL, they view that as a very high volume product and see the demand that comes out of there.

That combined with the cable modem will allows us to make some progress, but, again, of all the things going on, that's one I'm still concerned is going to hold us back a little bit from the ideal scenario.

MODERATOR: Great. Thank you very much for taking the time and thank you for your talk today.

(Applause.)