Innovation: Our Most Important Investment
Financial Analyst Meeting 2004
July 29, 2004

BILL GATES: Well, thanks, and good morning. I guess to kick things off talking about innovation and really give you a sense of how it fits into the software business model. We did achieve last week an interesting milestone in the business of software. I was out at a movie a few nights ago and some guy came up to me–well, a lot of people come up to me and say different things. I don't always pay attention. He said, “Thanks for the three bucks.” And I didn't know what the heck that was about. And then, at the end of the movie, somebody else came up and said, “Thanks for the three bucks.” And I realized they must be talking about the dividend. And being able to redistribute $30 billion all at once, assuming the shareholders approve of that, says that the software business model has come a long way.

BILL GATES: Good morning. I get to kick things off talking about innovation and really give you a sense of how it fits into the software business model. We did achieve last week an interesting milestone in the business of software. I was out at a movie a few nights ago and some guy came up to me–well, a lot of people come up to me and say different things. I don't always pay attention. He said, "Thanks for the three bucks." And I didn't know what the heck that was about. And then, at the end of the movie, someone else came up and said, "Thanks for the three bucks." And I realized they must be talking about the dividend. And being able to redistribute $30 billion all at once, assuming the shareholders approve of that, says that the software business model has come a long way.

It's interesting that throughout the history of the company, the use of capital has not been a limiting factor in our activities. That's not true of most businesses. Our businesses have really been limited only by our creativity in terms of new ideas, and our ability to put together world-class teams to execute on those ideas. And the fact that new ideas keep going is very, very important. One challenge in the software business model is that once you license a piece of software, it never wears out. And so if all you think of yourself as doing is basic word processing, or basic database, then at some point you saturate the customers out there, and you simply aren't charging forward, achieving new growth. In fact, your sales don't even maintain their current level because all you're getting is perhaps the maintenance from that base.

Now, Microsoft has always thought very broadly about software, whether it's thinking about the Windows franchise as a platform that needs to get richer and richer, or the Office platform as a franchise that needs to take all the productivity needs people have and embrace those. We are branching out into new areas, driving value into the software. The pace at which we do that, the pace of the innovation, really is a key variable in the financial equation. I've been surprised that many software companies don't broaden out, don't think of what they're doing in a comprehensive enough way.

In our case, we can think of any inefficiency—inefficiency in communications and information visibility, in IT costs, in development costs—as an opportunity for us to go in and say, hey, software can come in and make a very big difference. And so, as we drive forward with that, the equation about how many billions and more profitability we can generate will be determined by this. One of the factors that has been critical for us is that we get to run on an innovative hardware platform. If it wasn't for Moore's Law, that is, say we went a decade where microprocessors did not improve, that would be a big challenge to us because virtually all the ideas I'll talk about today are things that are being enabled by new hardware advances. Processor speed, memory space—we're going through a transition now from the 32-bit world to the 64-bit world, and amazingly this is going to be the smoothest, almost the quietest transition we've ever had in our address space activities. Moving from 8-bit PCs to 16-bit was a complete start-over. Moving from 8086 to 286 was very painful—extended memory, memory cards, all sorts of strange things showing up. And then going from that segmented model to the linear 386 model was also quite painful. A lot of starts and fits with OS 2 and how the model would work, how UNIX would take advantage of that … Today what we have is a natural extension that will be built into virtually all the x86 processors over the next couple of years. And so people who buy hardware will be able to run 32-bit OS and 64-bit OS. The 64-bit OS runs 32-bit software. And so slowly but surely, you'll see the applications recompiled for 64-bit, particularly things like database, where getting that extra memory size makes a huge difference in terms of the performance you can deliver. Networking performance is also important to us. If you think about how video across the network is becoming a very key scenario, that's only enabled because we have this incredible bandwidth. In fact, we'll talk about how TV will literally be redefined because the phone company, cable company and perhaps, over time, some wireless approach will deliver enough bandwidth into the home that having high-definition video feeds on demand, personalized advertising, that becomes a very practical scenario.

Storage capability, obviously moving at a pretty amazing pace, and we can have huge, huge mailboxes as part of free mail services, and even bigger mailboxes as people beef up the commercial mail servers driven by Exchange.

So everything here really drives our opportunity to come in with new software value. The idea that cameras are very cheap and that cameras will be built into cell phones means that for the first time the cell phone can basically see what's going on. It can take a picture of an expense report, recognize what's there, and send it up to an application. It can take a picture of a barcode, see what product you're looking at buying, tell what the price is, what the recommendations are—really engage in a very rich interaction between the physical world and the digital world because of the magic of software that is running behind that camera.

So we need these hardware advances. They're coming at a really excellent pace. Even broadband penetration, which has been kind of a limiting factor in terms of some of our consumer activities, is going up in a very healthy way. So nothing holding us back on this front.

If you look at the pace of software innovation, it's important to remember that we operate both in the business space and in the consumer space. The business space has been and will continue to be the biggest area of revenue and profitability for us. That is, the value, the efficiency for information workers worldwide, is something that if we can improve even a few percent a year, the economic impact and the ratio of the benefit to the cost of the software involved is really fantastic. We're talking about simply a few hundred dollars for a corporation to have the very latest productivity software for someone. And as a percentage of the investment there, that is very, very small.

I show here that we have a wide range of competitors. Many of them are not quite as focused on software as we are. Many of them that are in software are not quite as long term–focused, in terms of the R&D that they're doing, and thinking comprehensively about how the pieces fit together in a new way. But they are excellent competitors, they keep us very much on our toes. One thing I'm showing here is that the pace of new activity is actually somewhat higher—I show it as high-speed—in the consumer space than in the business space. And there are reasons for that, but they're reasons that we can change and will change. In the consumer space, upgrading to a new machine and connecting up to a new service, where the server that delivers it is running essentially in the cloud—the company that offers the service put that together. Moving up to something new is a lot more straightforward than bringing that server into a corporate environment, getting the hardware provision, the budget, getting people connected up with whatever software is needed there.

So we've seen, for example, in search, in music downloads, a lot of things are very rapid in the consumer space. Turnover of cell phones, expectations of new features there. The cell phone space is a good example of one where, if it had stayed as a fairly simply device, just voice and some contacts and things, we wouldn't have seen it as a huge opportunity; that is, the software value-add wouldn't have been that dramatic. But as we saw that over time, visual recognition; speech recognition; really rich calendaring; a rich understanding of what sort of notifications should interrupt you, depending on what shows up on your schedule, who it is; things of that nature are very deep software problems that we can bring into those devices, and those devices turn over at a rapid enough pace that calling that high-speed, I think, really makes sense there. The speed at which video gaming is going to move forward—you'll see over the next year some very dramatic things there. And so nothing holds back that pace, not relative to photos, not telephony, not new ways of thinking about networks, social networks merging with the idea of the personal home page, peer-to-peer communications—there's a lot we can do there. Very, very high volume–oriented, oriented toward subscriptions in some cases, advertising in some cases. So there's a lot we're doing there.

In the business space, why do I say that it's only a medium-speed pace of adopting new things? I think there are several reasons here. One is that having compatibility with what's existed before is very, very important. And that's especially true at Microsoft. When you look at our investment in product groups, you'll find that a very large percentage of that is the testing work that goes on to say that we can give you new capabilities, and the existing binaries will run, and they'll run in a very hardcore fashion. That's a big challenge, but it's something customers demand, and something they expect. That means that the pace that you can put new things in is not quite as rapid. The variety of applications that corporations have, the way they need to test them against new things, that acts as somewhat of a brake on adopting new versions.

One way to look at this is to think of a version tale. You can look at, say, ourselves in terms of our operating systems, or ourselves in terms of Office, and say, okay, what percentage of the users are back using previous versions? Obviously there's an upgrade opportunity for those people, but you've got to make the cost-benefit very, very straightforward, getting people up to the latest things, which is where you have the greatest feature improvements.

We can make that easier for people by giving them tools that allow them to test their software in more automatic ways. We can do that by letting them know all the different applications they have on the desktop, and move those often off in isolation to a server, to greatly simplify their desktop activities.

It's only in the last two years that the idea of updating the software on the desktop, without actually visiting the desktop, using primarily something we call Systems Management Server, SMS 2003, that that's done on a very automatic basis. And we have inventorying, and we've made that work even for portable systems. So a lot of what we need to do with software is increase the agility of the corporate server and the corporate desktop—really understand what are those things that hold people back, and get them in a strong position to always have the latest and greatest. In a sense that's a software opportunity for us, but it's of particular importance because it can gate how many of those innovations get through in a rapid way.

I've listed some of the competitors in that space; I think you'll hear more about the specifics of our competition in each of those cases. You can see what we do in terms of thinking about the information worker, thinking about all the professions in the company, all the different processes, and how productive they can be. We're in there at, in a sense, a point of integrated innovation that's very different than anyone else. Getting that departmental server, that desktop, those bottom-up applications, that rich visualization, the communications interface through Outlook, instant messaging—those are things that are very important assets for us. Because if people are used to working with our software that way, all the new innovations around document life cycle, protection, workflow, will come through the evolution of those interfaces and being compatible with the add-ons that work in those things. But we can do a lot to push things forward there. The security imperative we have, that I'll get into, makes that even more important: the idea of rapid delivery of improvements into the desktop.

Now, measuring innovation is not something that can be done with one simple number. Obviously, when Steve decides what level of R&D we're going to spend, that is a bet. And it's a very serious belief that the innovation investments will pay off in a very dramatic way. We've always been willing to do that in a very aggressive fashion, whether it's adding new businesses, or adding things within the businesses that we're in. One measure you can look at is the patents that we go out and apply for. And you see on this chart two ways of looking at that. The blue line shows the patents being issued from the Patent Office. Typically, now, if you send a patent in today it would be something on the order of three-and-a-half to four years before that patent would likely issue. Some are faster, some take even a little bit longer than that. If you go back three or four years, it was a little bit more rapid than that, probably more in the two-and-a-half to three years range. So you can see that's why if you shift the blue curve over a little more than three years it matches up to the green curve. And that's typically what you should expect. You have some patents that you file, which is the green curve there, that break off into multiple patents. Some of them, although a very small percentage, don't eventually result in the issuance of a patent. But you can think of pretty much a one-to-one correspondence between those two things with the time shift.

You see a couple of things here. You see the absolute number; that is, during this fiscal year we expect to file for a little over 3,000 patents. That means, starting this month and out through the next 11 months, a fairly dramatic increase over previous years. Last year we were at something over 2,000. Sometimes you'll see different ways of measuring this, where people count applications in different countries. The easiest way to look at it is just to look at U.S. issuance and U.S. filings, because there are very different practices from companies in terms of how many overseas jurisdictions they file in. Unlike copyright, to secure your rights you have to go into the various countries. And that's obviously somewhat of an investment. That's something you would also see if you looked at the international equivalent of this; you'd see the same type of ramp-up in terms of what we're doing.

It was probably 14 years or so ago when IBM, as part of their relationship with us, came and visited me and said they'd be willing to license their patents to us. And we said, oh, patents, wow, you want to license your patents to us. In fact, we did enter into an agreement with IBM, which IBM has done with many others, Microsoft has done with many others, where we had a certain type of cross-license. And I'd say year after year, certainly subsequent to that, it's something we've put a lot of energy into.

Patent numbers can sometimes be a bit misleading, because you really have a choice when you have an idea how many patents it breaks down into, and how novel or important that idea might be. Again, there's no precise science on this. This is one measure of patent importance called "current impact," and what it does is say, patents coming after yours, how much do they think your work is of enough importance that they cite that as prior art. And it's one of about four or five measures people use to look at overall patent quality. This is compared here to a group of other companies that do similar type work. You can see we measure up fairly well. Not a dramatic difference, ranging from 1.45 to 2.23. But we think patents are patents. What we're doing is, if anything, more valuable than what others are doing.

Our patents are on some fairly basic things: these new approaches to TV, or interactive gaming, the work we're doing in speech recognition and understanding. So we end up with a very broad portfolio, and that's something that we will continue to invest in and has always been something that we look at with other companies in the industry and decide what sort of licensing relationships we should have with them.

Just in the last year, examples of this are our cross-license with SAP; an IP license with Sun, as part of our new relationship with them. We have some cases like InterTrust—there was some level of court activity before we reached a license to their full portfolio that related to digital rights management. So a lot of activity, a lot of visibility.

The software industry is one that, other than the pharmaceutical industry, probably patents are the most important compared to many other industries. In the case of pharmaceuticals, it's the only thing between the generic and the actual recovery of investment by the inventor of the product. Here we have implementation, services, many other things that go into that value equation. But the patent piece is an important piece, and something that you might even say, industrywide, you see increased, intense focus on.

For us the cycle is pretty simple. We make the filings, that's securing our IP rights. We license these out to people—and there is quite a variety of licensees, little things like file formats, ways that we display text on the display, this thing called ClearType, we've got licensees for that. As part of licensing our products we include an indemnification, and that's something over the last few years customers have asked us for, and we've actually stepped that up quite a bit, to make it very clear that that's part of what comes with the software license. Then the licensing program, that's something that we're ramping up. Certainly over the years, as IBM ramped that up, they were able to do something pretty significant there. We're at an early stage on that, but it is something that we're pretty excited about.

Now let me turn to the concrete areas of research advances. The way we're organized, we have our product groups, and we have the dedicated Microsoft Research Group that we call MSR. MSR is where we have the things that are the most advanced. We don't have a schedule, we don't even know exactly what they'll come up with, but we have experts in all the different areas you'd expect: graphics, linguistics, security, databases. If you go to the various industry conferences—database, graphics, artificial intelligence—you can get kind of a measure of the amount of research going on in different organizations, just by looking at the number of papers at those conferences.

Actually, as we've invested in research, there's actually been a decline of the portion of papers at these conferences that are presented by commercial organizations. That is, as some of the investments of people like AT&T, Bell Labs, or Xerox, or some of the labs that were established here, actually the number of papers they're doing is going down relatively, and in absolute, and of course universities make up the bulk of the rest, and that's continued at a very strong pace. So our absolute is going up, and our relative portion of the commercial papers and activities there is actually surprisingly high.

You'd think, given that R&D has this big payoff, that you'd see increased investment in it, but in fact, surprisingly, we're not seeing that in such a dramatic way in terms of other U.S. software companies in particular.

Now the Microsoft Research Group, we try to make sure that it's not separated from the product groups. Certainly examples in the past like Xerox, where the quality of the research wasn't fully paid off in terms of how it was used in their products, was cautionary to us as we started making investments here. And our Microsoft Research Group is in Beijing, in China; here at our headquarters; and the other large site is in England on the Cambridge University campus. We also have a smaller site down in California, with some very focused researchers there. Speech is an area we've been working on for many, many years. And this is the first year we've actually released a product with speech-serving capabilities. Here, as you look at limited vocabularies, as you look at more powerful environments, the role of speech will just increase over time. The holy grail of total recognition at human levels is still many years ahead of us, but the kind of breakthroughs that are necessary are taking place, and it's very practical for a lot of things, even now.

Ink and Tablet: Ink recognition is something that we've said we're going to make this a very mainstream thing. We're off to a very good start on the Tablet, and we're going to show you the latest generation of Windows that's actually coming out this month, along with SP2 that implements that.

The spam problem: I'm sure you're all inundated with spam, or at least you have been. Spam delivered to users on things like Exchange and Hotmail has been cut down in absolutes. So if you're on a mail server where the Microsoft Research spam filters are there looking at spam patterns, that's cutting it down in a very big way. If you have a spam recognizer that's kind of hard coded, the people who generate the spam just watch, generate new things, and they get around whatever your specific word checks might be. The only way to make it tough for them is to do what is called "machine learning." That is, have a base of what's not spam and what is spam, and then have software go through and look at that and say, okay, what word frequencies are different? What kind of Web sites are linked to? What kind of imagery might be connected with this? And by doing that you make sure that it's actually resilient as they change their techniques, you see how they're changing their techniques, and it adapts to that.

We are going to complement this with some other Microsoft Research work that has to do with somebody who is intent on sending communications to you either being verified that they are on your Contact list, or being verified that they're willing to give up some computation time on their machine to prove to you that they really want your time, and to do that in such a way that it's not economic for somebody who is sending out millions of pieces of e-mail that have a very low response rate. That is, actually put some friction in the system so that the worst kind of spam becomes uneconomic.

The search area: I hope some of you have tried out the beta we have of our back-end search engine. It's really been amazing how quickly we've pulled together the work there, and a huge amount of what we've been able to do is based on Microsoft Research. In fact, many of the things where we will not only match but exceed the state of the art are based on those things. And finally, I mentioned work-around notifications. If you've got Outlook 2003, you know how it comes in with a little alert box but that that fades away. And a lot more to come in that notification-type capability.

So we've got two quick demos we wanted to show you of these things that are actually either in the marketplace or about to get into the marketplace. So to start that off, let me ask Cory Linton from the Tablet PC group to give us a look at the next version.

CORY LINTON: Thanks, Bill.

Tablet PCs launched about 18 months ago. Since then, we've seen second-, third-, even fourth-generation machines. We have new OEMs. We've seen prices come down. In a few weeks, we'll be releasing the new version of our operating system, Windows XP Tablet PC Edition 2005, that will ship as a free upgrade with Windows XP SP2.

I would like to show you some of the power of that operating system, and how it really enhances the ink-to-text experience. To do that, let me just open up a new e-mail as if I was writing it. For those of you who have used Tablets previously, the way you enter text is, you go down to the taskbar and you push a button, and you write the text down at the bottom here, and then it goes into a text box up above. And we found with user research that users just didn't like moving their hand up and down. The screen is just too big; it becomes a real tedious process.

So, what we've done in this version is we have what we call an in-place tip, and this is the ability, if I click on any text box in any existing Windows application, I get a little orange button, and I can write right in here. Previously, this was only available if an application wrote some custom code. Now it's available for all Windows applications.

To further show you the power of this, let me write something here. And you'll notice a couple of things. It's automatically growing. I can just keep on writing, and it will add new lines as I need. And it also gives me real-time feedback on what I've written. And you can see the recognition actually was very good, but it gives me real-time feedback. If I have a mistake, I can click on a word, and I could actually just change those letters. So I can fix it like that, and I can insert the text. So it's a much easier experience to enter text in applications. The keyboard is still there; if you need to write long documents, you'll still use the keyboard, but the basic text input experience is much improved.

There's another concept in this new version that we call "context awareness." And this is the ability for the recognizer to know what kind of data to expect. And Bill alluded to this as he was talking about speech. So if we can tell the recognizer that whatever goes in this field comes from this list of words, the recognition accuracy goes way up. We've actually applied this to Outlook, and to a state field here. So I can scrawl in "California" pretty sloppily, and it's choosing from only 50 words, so its chances of being correct are really very high. What's neat about the context awareness is that companies will have a tool called the Context Tagging Tool, and this allows companies to open up an application and say, this is a state field, this is a ZIP code, this is a customer name field that comes from our list of customers, this is a product ID field, without having to crack any code. So organizations can take a form and make it optimized for the Tablet without having to change any code. So the millions of forms out there that are used in mobile situations can now be easily upgraded to run very well on the Tablet PC. So that's basically our OS.

I want to show you a couple of works of innovation that have been done by outside parties on our platform, because the other big investment in Tablet PC is having a very rich platform that developers can ride on. The first one I'm going to show you is kind of a fun consumer tool, and this was done by a developer in a contest we had to create the coolest Power Toy. And this is called the My Own Font Tool. And it allows you to write a font in your own handwriting. I'm not going to fill all this in. I'll open a project I've already created of my own very stylized writing. And what's neat is that I can create this font, and then I can apply it to the system. And when I do that, my own handwriting is suddenly throughout Windows. And it's really the ultimate personalization of the machine. And you'll see here that as this thing gets applied, it looks very different, and it feels very much like my machine. So I can pull up the Start menu and suddenly it looks really different. And nobody can read this, but I don't care; it's my machine, and it's my handwriting. I can pull up applications and the menus are all in my own handwriting. And of course, I can open an e-mail and type an e-mail that looks like my own handwriting. So you can really see some fun innovations there.

Now I'm going to show you an example of some research work that's been done by Brown University in the area of math recognition. And this is a project to explore how the Tablet can enhance math learning and make it easier for students to grasp the difficult concepts of math. And what you can do is write an equation and have it recognized and do things with it. So let me open one that's a little more interesting. This is a pretty complex equation. Once I write that and recognize it, I just draw a line like this, and it's going to graph it. And I can change the parameters. I can write -7 to 4, update it, and it will change the graph. Many of you have long days in finance studying pretty complex equations. Imagine if you had a tool like this that visualizes what's happening, and you can change things in real time. It really becomes a very powerful learning experience.

Let me open up one that's a little more complex. This is an equation that represents a spring, and the Y of T equation here is the movement of a spring. I'm going to draw a spring here, I'll draw a little weight, and then I connect the weight to the spring, and connect the spring to this line, and then I draw a line through these equations and it binds them together. It says that the B in the Y of T equation comes from this B of 3, and then I'm going to touch that weight, and I've now told it that that weight is going to be driven by that equation. And this is just pure math. There's no tricks or physics involved. And when I hit Run, I see the spring move. So imagine the power of learning: I can play with the variables, see how it affects the spring. It really becomes a very powerful thing.

So I've shown you a sampling of our new operating system, Windows XP Tablet PC Edition 2005, that will ship in a few weeks with Windows XP SP2 as a free upgrade, and some of the neat enhancements on the platform.

To show you some of the great work that's been done in natural interfaces, I want to introduce Kevin Shaughnessy from our Speech Server team.

KEVIN SHAUGHNESSY: Thanks, Cory.

So as Cory demonstrated, our research and innovation in pattern recognition technology is finding its way into shipping products like Tablet PC in the form of handwriting recognition, and the same technology has also influenced our efforts in advanced speech recognition.

So before I demonstrate this, though, what I'd like to do is quickly set the stage for the business value that speech recognition technologies can provide.

Today we all know that Web technologies provide great benefits to enterprises by offering automated self-service solutions to customers and employees around the world. So, for example, I have a Web application here, which is going to appear on the screen as we see it, and this Web application is a Web portal. It's a fictitious Web application for the Grand Deluxe Hotel, and with it I can book a room in, say, Chicago, Illinois, for three nights starting tomorrow. And here I'll choose from one of the three available room types. I'll choose the lower-cost economy single option and I'll log in as our demo persona, Joe Lucas, and we'll use Joe's stored profile information to book this room, and we're done. The room has been booked and the confirmation has been sent to Joe's e-mail.

Now there's nothing really new about this, and today we're all very familiar with this type of Web application. And applications like this have significantly reduced the number of expensive and routine customer phone calls coming into company call centers.

Yet customers don't always have access to the Web, so companies continue to receive millions of phone calls a year. Well, now these companies can further reduce costs in their call centers by automating these telephone transactions using speech technologies.

Let me demonstrate. A moment ago I booked an economy single room for Joe using the Web site. Now let's make a change to that and let's change that to an economy double room, but this time let's use the telephone to do so, to access the same Web application, but now speech-enabled and running live on Speech Server.

COMPUTER OPERATOR: Hello and welcome to the Grand Deluxe Hotel. Would you like to check room availability, contact a hotel guest, or change an existing reservation?

KEVIN SHAUGHNESSY: I'd like to update my reservation.

COMPUTER OPERATOR: To change an existing reservation, please key in your preferred guest membership number or just speak your name.

KEVIN SHAUGHNESSY: Joe Lucas.

COMPUTER OPERATOR: Okay, I found a reservation for Joe Lucas. For verification, what is your billing ZIP code?

KEVIN SHAUGHNESSY: 45245.

COMPUTER OPERATOR: Great. Your current reservation is for an economy single for three nights starting tomorrow. What would you like to change?

KEVIN SHAUGHNESSY: I'd like an economy double room.

COMPUTER OPERATOR: So you'd like to change from an economy single to an economy double for three nights. Is that right?

KEVIN SHAUGHNESSY: Yes.

COMPUTER OPERATOR: Okay, an economy double room is available for three nights starting tomorrow. Would you like to make this change?

KEVIN SHAUGHNESSY: Yes.

COMPUTER OPERATOR: Thank you. Your reservation has been changed and a confirmation has been sent to your e-mail address. Can I help you with anything else?

KEVIN SHAUGHNESSY: No, that will do it.

COMPUTER OPERATOR: Thank you for calling the Grand Deluxe Hotel. Goodbye.

KEVIN SHAUGHNESSY: So this was the same Web application that we saw earlier, but now speech-enabled for access over the telephone. And this is all running on the same Web infrastructure, it's using the same back-end data and it's also built on the same code base. But the value of speech technologies in the enterprise doesn't stop there. Companies can also add speech to their Web-based mobile device solutions. For example, the Grand Deluxe Hotel has given its employees Pocket PCs similar to this one, wirelessly connected to their onsite intranet. This allows employees to better assist customers from wherever they are in the hotel.

So you can imagine it's a few days later, and Joe is now staying at the hotel, and he wants to update his room yet again, this time to a deluxe suite. I'll play the role of a hotel employee, and to assist Joe I'll use this Pocket PC and speech.

View current guest record for Joe Lucas. (Spoken into device.)

So with a simple voice command, up pops Joe's current information, and I verify it with him, and now I'll check for available suites.

View all available suites. (Spoken into device.)

Here I can see there are two suites available, and I'll get a closer look at either one by simply tapping with the stylus. And now to move Joe to this suite, I'll use another voice command. Update current guest record with this one. (Spoken into device.)

I confirm the change and now we can see that Joe has been updated to this deluxe suite for the remainder of his stay.

What we have here is a unified, speech-enabled Web application that's accessible from multiple touch points, including the Web, the telephone, and mobile devices. Just as companies like Rite-Aid and Talbot's and GMAC are already benefiting from these types of speech-enabled solutions, you could imagine how this technology can also benefit a broad range of enterprises in helping them to further reduce costs, increase employee productivity, and enhance customer satisfaction.

Thank you.

BILL GATES: Thanks, Kevin.

Now let me look at some of the future things that research advances will do in software, things that have not been done before. If you imagine in this digital environment, what would you like to have when your kids are growing up? What type of photos, audio recording, movies, that would let you go back and capture those memories whenever you want? How would you like to be able to navigate those things—remember the movies you watched together, the books you read together, call those things up and make it simple to navigate based on whatever you're interested in? That's a software opportunity that isn't just about photos or movies, it's about the structure of the information. It's about making it very easy to gather it and automatically recognizing things so that you can move around in that information.

Digital media: We'll show you another demonstration that will give you a glimpse of where we're going with the "Longhorn" user interface, what type of navigation of media that having rich storage and a unified approach will give us, and again this comes out of our research work. Modeling is something you're going to be hearing more and more from us about. Modeling is the idea of declaring what something should do, whether it's a business process or a computer system, and having the rules and constraints for that behavior and expecting the software to figure out how to do that without being told in such an explicit way.

It's modeling that's going to greatly simplify applications customization. It's modeling that's going to let navigation of business information be done in a richer way. So visualizing models, having rules in models, having rules that talk about workflow and communications activities, those are things that there's a base set of technologies around modeling approaches.

The place we're using it the most today is to describe how two pieces of software relate to each other. If we want to test a large piece of software, the best way to do it is to be able to test individual components. But in order to do that, you have to understand all of the rich ways that other software calls into those components, and there's never been a way of modeling that. We created a rich new specification language called Spec# that lets us do that, and then we go in and prove that given those specifications, this will do what we expect it to do.

These kinds of proof techniques have been talked about in academia for many decades, but now they're actually in use today, making sure these systems are secure and reliable. It's a case where we actually scaled up these techniques and made them practical. That was a first, and it was really kind of a necessary thing to deal with the scale of software we have and the kind of guarantees people expect, particularly around security. Large displays are becoming commonplace, and yet the user interface that we have today doesn't automatically work that well. The cursor gets lost, too many windows, you can't group things together, you don't know where you're going, and so the experimentation around that has really come up with a lot of very interesting ideas: dropped targets and moving the icons around so that you don't have anything that's docked to the bottom.

Document intelligence is being able to take any document, even an old document, and recognize exactly what's in it, what are the high-level constructs there, and therefore what are the parts that might be relevant? It's a lot better than the kind of text indexing that's being used in search technology today. It's really parsing the sentence and understanding what's going on and seeing all the different pieces, the diagrams, the images, the annotations, and making sense of those semantically, at a very high level.

I mentioned vision as something that's pretty exciting. You ought to be able to look at photos and tell who's in that photo. If you have a camera on your PC, and you're doing something like a Live Meeting connection, the fact is that as you move around it's not very good. You don't appear to be looking at the camera; it's a very poor experience. But if we have dual cameras with software that's recognizing how you're moving, we can actually focus that in the right way. In fact, if we have a room with lots of people in the meeting, we can see exactly who the speaker is, create a view of the speaker as well as a panoramic view that adjusts for the angles in the room, and make that remote experience dramatically better.

So vision is coming in as a central piece of the platform, built into the PC, built into the cell phone, built into little peripherals, in a very rich way.

Finally, there's the idea of mesh networking. This is the idea that if you have wireless capability, why can't you have a group of people essentially have hops between each other to create the network, rather than having to run as much cable or fiber to create that network? Particularly in developing countries, this idea is very attractive because communications costs are really the thing holding them back. It's not the hardware and software costs, it's that difficulty of getting the high-speed infrastructure out over a large geography, and mesh networking promises to help deal with that as we get wireless advances like WiMAX, and better and better Wi-Fi capabilities.

So to give you a glimpse of what you'll get in the area of digital media, I'd like to ask Steven Drucker to come out and show us the media browser.

STEVEN DRUCKER: Thanks, Bill.

BILL GATES: Welcome.

STEVEN DRUCKER: I'm going to show a quick demo of a research prototype for managing digital pictures. People today are taking hundreds if not thousands of pictures, and this prototype explores some new ways in which we can make that experience a little bit better. So first of all, one of the things that we can do is exploit the graphics capabilities that you're seeing being used in video games today to quickly be able to zoom in on these pictures. I can pick a picture, show the full size of the picture, and you should also be able to manage digital video because your still cameras are taking video today, and vice versa, so I can very quickly be able to show video.

Well, if that's the case, if you're taking digital stills and digital video, you should have a quick way of filtering out and just showing the things that you want to show. So, for instance, I can move up here, I can select video, and I can quickly filter down and the images expand on the screen to use the maximum allowable space so you can see them better.

Well, digital video, digital stills, it's pretty easy to tell what the difference is because the camera marks that. But we also have the vision analysis routines that Bill was just talking about, so we can do things like face detection. I can click up here, I can detect all the pictures with faces in them very quickly. It's face detection, not recognition, but still it really helps out finding those pictures, even if they're tilted.

What fools Mother Nature also fools the face detector. But we choose to err on the side of inclusion rather than exclusion because, really, the point about this whole system is to be able to help you annotate the pictures.

We have a couple other features. We can detect indoor and outdoor shots. Here it's looking for horizon lines and colors, so it's looking for horizon lines and blue, greens, et cetera. But really the whole point, as I mentioned before, is to be able to annotate your pictures with keywords.

So first I'll show you, if you have those keyword annotations, how you can use them to quickly find the pictures you're looking for, and then I'll show you how you assign those keywords. So here we can just again move down these keywords; you can see them highlighted on the right. If I choose to, I can say "computer museum," let's filter down to computer museum, and I just show the shots in the computer museum, and now I just show you the keywords for narrowing that down further. So I can say "Gordon Bell at the computer museum" or even "my favorite of Gordon Bell." This is a sophisticated SQL query that we've constructed on the fly, through a convenient user interface, people really can see that by doing incremental improvements to the query.

So how did I assign those keywords in the first place? I mean, I could sit here and assign a keyword one at a time and that would be insane. No one in their right mind does that. What we can do, though, is we can make use of the only metadata that we start out with, the only information on those pictures, which is the date and time that the pictures were taken, and then we can use the computer to cluster them and present that to the person, the user, in a better fashion.

So here we've clustered the pictures, and now the person can look through and say, oh, okay, these were all the pictures I took, then I put the camera down and it made a break of a column. And I can zoom in at any point in time and recluster it and say, yes, yes, I understand this, this is exactly the day that I went to the Mariners-Red Sox game, I can select all of these at once. I can assign a keyword to all of them at the same time. And that really makes it easier, because as you're browsing you can find the clusters, you can find the patterns.

There are a couple other ways that we can select many at the same time. Also by date you can say, okay, I want to select right here, and I want to expand it outward. I want to say, show me everything within four days, within a month, within a year, and you can keep on searching for things this way. I'm going to start using these techniques in conjunction with each other, so now I'm going to filter them down.

I can show similar-looking photos. So, for instance, if I like sunsets, I can find that. If I'm interested in bridges, I can find other bridges, and it matches other bridge shots so I can then label those all as bridge shots.

So one of the things I'm going to do now is I'm going to start pulling out some pictures of Thanksgiving, because I'm going to want to send these around to people. I may want to send them via e-mail or other devices. But I don't want to send everything; I want to start using all those tools in conjunction. I'm really interested in sending this to my friends that were at Thanksgiving.

So I can go up here, and I can pull up the face detection, click, and I can remove all those pictures that did not have faces in them.

Now I can send these pictures to devices in our house. In fact, Curtis Wong is going to show you what happens if I send this to a Pocket PC.

CURTIS WONG: So what I have here is an ordinary Pocket PC. We've written some software to enable this to emulate as if it was a wireless picture frame.

I don't know if you're like me, but I have a few thousand digital pictures on my computer, most of which I never really see. But imagine if you had a wireless picture frame in your living room and that 24/7 those pictures were always appearing there using the intelligence of the PC to pick the appropriate pictures to show from your entire collection. So you could set simple preferences, such as same time of year, so that when your birthdays and holidays appear you start seeing those pictures from years past of those particular events.

So what we have here is a variety of pictures, and I can go backward if I missed one, or see one here. I can also go and pivot to show related pictures, and it will bring up pictures that are related to this. Now you remember the simple annotation that Steven did before. It allowed me to pivot based on other pictures that we've also assigned keywords to. So in essence, every picture that appears on this frame is a point of entry to potentially everything in my collection.

I can also record a voice annotation. So let me go here, and I'll hear and record something. So say I'm sitting here with my mom, and we're looking at pictures, and I say, "Hey, Mom, what do you think of this picture?" And she'll say, "Oh, that's such a cute picture of Laura." And that little annotation is immediately accessible and part of the database with your pictures, so that seamlessly and easily if you go and look at your pictures in the future, that same annotation is there. STEVEN DRUCKER: What you've seen right now is that we're trying to explore—we're working with the product groups to show how we can exploit the graphics capabilities of the machine, how we can use advanced vision algorithms to do better metadata annotation, and how to use this seamless computing experience all to essentially integrate everything and do the appropriate things on the appropriate devices. And we're working with the product groups to make these experiences available in Longhorn and beyond.

Thanks very much.

BILL GATES: Great. Well, one thing about innovation is we've got to make sure it actually maps to what a customer sees as value. So let me touch quickly now on how we see ourselves reaching out with customers really as their advisor, their partner, to help them solve some problems that software technology can make a big difference in.

First and foremost for us now in our dialogue with business customers is their costs, their complexity, their concerns about security and management. They want very clear guidance on how they isolate their networks, so they won't be subject to attacks. They want to understand how to update, which updates to flow through very rapidly, which ones can wait longer and go through a complex testing cycle. They want to understand how to move away from passwords to smart card or biometric systems that will secure their corporate information in a much better way. And today what goes on behind the scenes, in terms of setting up group names and authorizations and monitoring systems, it's too complex. It's too manual. There are too many moving pieces there that need to be built into the system.

And so our promise to them is that we can actually take costs out at the same time as we improve the reliability and the security of these systems. In many cases we're simplifying their software environment, because the richness of what we're now delivering allows them to count on us for a higher percentage of what they need.

The advances we've seen with System Management Server, with MOM (Microsoft Operations Manager), our ISA product, and many new things to come into this space, should make the security area something that turns from a concern for us into something that's a significant, unique asset as well as a business opportunity, a lot of this built into the product but some of it coming as add-on capabilities.

The productivity area is probably the most important franchise that we have. Office defines business productivity. And yet we see Office in such a wide way that we really feel like we're not even halfway toward what we can do.

Collaboration: We took a big step forward with SharePoint, very important, we're getting good take-up on that. A lot of good stories you'll hear with some of those from Steve Sinofsky later. Business analysis: Still very far short of what people should expect, in seeing their quality results or their customer feedback or sales analysis. And we want to really be this trusted advisor who tells companies, for salespeople, for engineers, for marketing people, how projects get tracked, how things can be done in a much more effective way.

So all the different pieces, Exchange, which has gained a lot of share recently, and SharePoint, which has really taken off in a big way, and those new Office advances go together in that arena. When we think of the opportunities, they're pretty broad. Workflow is this idea of easily setting up a sequence of things. Even basic scheduling today is not a mainstream activity.

Unified storage says you don't want mail to be separate from your files and other things; you don't want the way you organize on the server and the client to be different. And so, particularly with the new platform technologies, there's a lot we can do there. We make sure that our Office group is taking full advantage of the platform, and that they really will guide the way for how those technologies drive real value.

As a development partner, we've been THE company in the industry pushing a service-oriented approach, the Web services approach, and we've been very gratified to have strong partnerships with people like IBM in defining the technical standards, and to see the uptake of these capabilities.

It was a big bet for us to build our tools around this, to design .NET to be the best Web services platform.

A huge milestone for that that you'll hear more about is the release that comes out next year of Visual Studio. That is how people can take the basic SharePoint capabilities that are built in and customize them for their needs, build Web parts that show the information in their company, so-called portal type capabilities—customize those and have a development story that ranges from just simple Office customization all the way up to the most complex enterprise applications. We want to span that whole space.

And our new work really moves us up on the high end. We decided to do green field development to have these deep, architectural tools. We decided that the right approach was to build something that hadn't been done before, and yet connect that up to our traditional strength down in Excel macros and Access, and have it fit in with the simple approach around Web services.

The application group is obviously a big investment for us in terms of the companies that have come together to form Microsoft Business Solutions. Particularly in medium and small business, this idea that we can have a comprehensive solution, some of which is software running on premise, some of which is us being willing to monitor those things through services that we provide, giving them the exact mix of that that they feel comfortable with. That's a very strong strategy for us.

We've seen with things like Live Meeting that software as a service, software running remotely, is part of the mix, not in the simplistic way people talked about three or four years ago, but now with the right kind of security and customization, that's coming back into the mix and will be very important.

Consumer space: A lot going on there, whether it's the mobile devices, MSN. You'll hear about our ambitions in search, and they are very exciting ambitions.

TV: I mentioned being redefined by the kind of bandwidth that next-generation networks will provide, not just digital video recording—although we believe in that, and will have ways that Microsoft products from Media Center and several others can offer that—but the entire network letting you have high-definition video when you want it, blurring the boundary between what's an interactive entertainment experience versus what's a passive video-watching experience, and really connecting that up to the PC. Not saying, let's just have dedicated devices, but rather the richness of the PC connected up to all the screens in the household and so being a unified store for music, photos, videos and those things.

We think Xbox Live is pointing the way to redefine gaming. A lot of investment there, but something that we're very excited that we're part of, because if you think about the home, entertainment will be a key focal point, along with the productivity and communications capabilities. And we are the company that has all three of those legs. We knew that the entertainment leg would be very, very important to have that kind of consumer solution. And then, finally, I talked about the software opportunities that our research advances bring us, in terms of letting people deal with their memories in rich and interesting ways.

So we've had a lot of advances. I mentioned the consumer space, and I thought I'd show you one of the recent advances that's definitely kind of consumer-oriented.

(Video segment.)

BILL GATES: So obviously the toaster is not the key thing we want to connect up there, but making all of that automatic through the plug-and-play infrastructure is something that the next generation of Windows does come out with, and I think is pretty important.

In the home environment, in a sense, the software has to be even better because there's no IT department to connect up to. And so the trajectory we have of making things far more automatic for IT departments, even in very complex situations, that kind of model-based approach, automatic recognition of what's going on with the system, that's very, very important for the consumer market as well.

So a simple summary here: We've been telling the same story, in a sense, ever since we were a public company. We believe in software, software as a business. We've generated significant profitability. We think there's more profitability where that came from. It only comes, and even the revenues we have all come through breakthrough innovation, and that's why the investment levels are up, that's why the focus on intellectual property is very strong, that's why making sure we bridge from the technology to what the customers expect, how they see the value out of that, is very important. And I think what we've got in the pipeline now is probably more exciting than what we've ever had, so you can expect big breakthroughs in the years ahead. Thank you.

(Applause.)

END Due to the varying sound quality and subject matter of tapes, the information in this transcript may contain inaccuracies.

Due to the varying sound quality and subject matter of tapes, the information in this transcript may contain inaccuracies.