From  http://www.microsoft.com/billgates/speeches/2001/08-07aiconference.asp

            Remarks by Bill Gates
            International Joint Conference on Artificial
	    Intelligence
	    Seattle, Wash., August 7, 2001


            Today, we can celebrate the fact that the PC is out in
            the hands of over 500 million users. In fact, this week
            I'll be attending a 20th anniversary event for the PC,
            and some of the pioneers will be getting together down
            in San Jose and talking about what has happened during
            the last 20 years, is that different than what we
            expected and how far have we come.

            My view is that it's definitely a glass half full. It's
            fantastic that we've changed word processing. It's
            fantastic that when I get up before you in a
            presentation like this I don't have any slides. I don't
            have to think about whether they're upside down and
            backwards or whether they're going to fog up. Some of
            you here are young enough that you can't relate to why
            that was such a problem in its day -- (laughter) -- but
            believe me it was. And so some very neat things have
            happened.

            The idea that computing is for the masses; that is, as
            an individual tool is very different than 25 years ago
            when it was more about big computers that companies
            would use to bill people and track information in ways
            that didn't seem very attractive.

            But we need to go a lot further. Certainly, today the
            amount of confusion and frustration that exists as
            people are trying to communicate, use their PC as a
            communication tool, try and use it as a creativity
            tool, there are serious problems there.

            And so for us there's a need for a much more natural
            interface. There's a need to model what's going on with
            the user. We have done some things where we've taken
            Bayesian models and use those in what we call
            troubleshooters, so if you can't print or your system's
            not doing anything at all, we go through and ask you a
            series of questions. And those have helped a lot, but
            we can see building a much richer model as being key to
            having the PC fulfill its promise.

            So the vision of Microsoft is pretty simple. It changed
            a couple years ago. For the first 25 years of the
            company, it was a personal computer on every desk and
            in every home. And it was a very good vision; very rare
            for a company to be able to stick with something like
            that for 25 years. The reason we changed it was simply
            that it became acceptable. It wasn't wild. It wasn't
            this big claim, where people would say, "Are you
            kidding'" You know, they would kind of say, "Of course.
            What's next' Is it some other company that's going to
            drive the next revolution here'"

            And so as we stepped back and looked at what we were
            trying to do with the programming model, turning the
            Internet into the fabric for distributed computing,
            getting your information to replicate in a very
            invisible way so that it was available to you
            everywhere, thinking of this programming model spanning
            all the different devices, we changed to the mission
            statement we have now, which is empowering people
            through great software anytime, any place and on any
            device.

...
            We say that we're at the start of a new phase that
            builds on all the previous ones, just like has happened
            before, but it's different in that there is a key set
            of standards and a key set of tools, a key set of
            applications that define this era. The standards around
            XML and the distributed computing protocol SOAP, and
            the related standards that come out of that, the idea
            that you can have the intelligence both be centralized
            and on the local device, that you can represent
            heterogeneous information, there's a set of development
            tools around that. The key scenarios are e-commerce and
            very rich reformed communication and those are the
            things that will drive this to be even more popular
            than the PC itself has been up till now.

            The scenarios that are very well established are really
            productivity and e-mail. Even in productivity we see
            new horizons. For example, if you think of the
            spreadsheet today, the way that information is
            represented there is extremely low-level. Yes, we
            understand the equation but we don't understand the
            interrelationships of the various things. If things are
            variable size, they don't get represented very well. If
            you imagine doing the forecast where you have to go out
            to different Web sites, it's a very manual process.

            And so we drive our new work in terms of scenarios. I
            won't go into all the different ones here, but I just
            want to touch on a couple to give you a sense of how we
            start with the idea of this being a tool and then we
            look at what kind of technology will allow that to be
            possible.

            Digital reading is a scenario we've believed in for a
            long time. Again, this is something where it seems like
            common sense, the fact that you could read off of a
            screen, you could annotate it to share it with other
            people, sort of the original hypertext vision, you
            could go back and search things that you've seen
            before, you can get it totally up to date, you can
            traverse a link to get in-depth information. Digital
            reading should be superior to reading off of paper.
            Well, of course, there are some roadblocks in terms of
            the size of the device and can you hold it in your hand
            and the resolution and just simply comfort with this
            approach that have not yet been overcome, and we see
            over the next two or three years that's definitely
            going to happen, whether it's the Tablet PC form
            factor, the wireless network, and so we see that
            there's a lot that we can do to make digital reading a
            much richer experience than paper-based reading has
            been.

            In the area of meetings, we think that facilitation of
            meetings both before the meeting takes place, during
            the meeting and after the meeting, there is dramatic
            opportunity. Some of this involves having a camera,
            which makes digital recordings of things that are going
            on, understands who's speaking, ideally building a
            transcript through speech recognition and letting
            anybody who wasn't present do the kind of searches they
            might be interested in, whether it's just looking at
            the transcript or seeing sort of a sped-up video that
            fills them in on the parts that might be relevant to
            them.

            In the living room, which is a tough frontier, we're
            taking a product we call the Xbox, which brings a very
            reasonable level of graphics capability and hard-disk
            storage for the first time into that environment, so
            that the kind of online games we can have through the
            rich hardware and the Ethernet connection that's built
            there, ought to push things up to a whole new level.

            Now, behind those games, we need very rich models. We
            need things that people find entertaining and rich, so
            that they'll go back week after week, month after month
            and make that something very real.

            So every one of the scenarios here, whether it's simply
            helping the user out or making it as rich as they would
            like it to be, involve some technology that would be
            thought of as artificial intelligence: looking at the
            text of mail messages to help somebody categorize them
            so that everybody doesn't end up being a mail clerk
            setting up their own e-mail folders. Things like that
            are we think within reach and drawing on the rich work
            that people here are doing.

            Now, one thing that is very exciting is we don't see
            any limitation on the hardware side. That's not to say
            that we've got infinitely free computing, but as we
            think about the processor speeds, the disk storage
            sizes, the kind of peripherals we'll have in terms of
            still image and video capture, the kind of microphones
            that will standard built into the PC, the array
            microphones, we see the hardware as being there for
            these advanced scenarios. High-resolution LCDs, the 200
            DPI LCD, which, when combined with our ClearType
            approach, gives you incredible readability; that is
            going to be a relatively cheap device within the next
            few years. The miniaturization to actually have the
            tablet PC be under two pounds and something that you
            would take to a meeting and simply do your notes on,
            that's absolutely within reach.

            The only piece of the picture where the hardware and
            communications people are letting us down a little bit
            is broadband connectivity out to homes. To businesses
            it will be there, but to homes even five years from
            now, even in the United States we're looking at
            something like a third of homes having the connection
            and two-thirds not having that connection. And so there
            is some compromise there in terms of what you can do
            when it's not always on, when you have those bandwidth
            limitations.

            But with that one footnote, a significant footnote, the
            same assumption we made when Microsoft was founded that
            we could focus on software and that there would be lots
            of new horizons that Moore's Law-type advances in
            hardware would be opening up, that's stayed true to
            form and certainly for the next decade that's an
            important thing that we'll be able to take advantage
            of.
...
            Fortunately, the market for software allows us to
            pursue these projects despite the large commercial
            scale and the cost involved in that. If you take our
            R&D budget, which is heavily development oriented
            percentage-wise, it's over $5 billion for this year.
            And when you have an image of that in your head, you
            don't have to think about big fabs or any capital
            equipment. That money is programmers sitting in their
            offices writing code. It's 100 percent personnel
            related. There's nothing in a software factory except
            people actually doing that work.

            One aspect of our R&D is the Microsoft Research, and
            that's been a wonderful experience. You probably know
            we're located in Beijing, Cambridge and here at our
            headquarters. We've been able to bring in an incredible
            range of great people and had those people reach out
            particularly to universities and other corporate labs
            and collaborate on some very advanced things.

            And if people want to know what we believe in and where
            we're going, they can just look at our Microsoft
            Research Web Site and they'll see the kinds of
            frontiers that we think software is pursuing.

            In the vast majority of that Microsoft Research work,
            areas that fit within AI are central to what we're
            doing, whether it's decision-making learning, language,
            speech recognition; these are the classic goals of
            artificial intelligence. We are putting our money where
            our beliefs are that these things will become real and
            allow us to build far, far better software products
            than we have today; and not far better for small
            audiences. We're talking about software products that
            many hundreds of millions, if not billions of people
            will be using and taking advantage of every day, things
            as simple as -- and we'll get into a little demo of
            this -- if I want to communicate with someone, how do I
            make sure my time is being used in the best way. It's a
            huge problem today and it's simple to state but very
            hard to solve.

   ...