BMOC #1: Bill Gates

College Tour
Carnegie Mellon University
Bill Gates
Thursday, February 21, 2008

JARED COHON: Good afternoon. My name is not Bill Gates. I'm Jared Cohon. I'm the President of Carnegie Mellon University, and it's my great pleasure to welcome you to this very, very special event. I'm proud of Carnegie Mellon. I'm always proud of Carnegie Mellon, but I'm especially proud of you today. The enthusiasm that you've shown for this event for Bill Gates is phenomenal. Your patience and perseverance in getting into this place are both notable. And it should be pointed out, there are 720 seats in this room, and Maconomy is also full. Hello, Maconomy. I can't hear them down there. So, by my math, that makes about 1,200 people who are sitting down waiting to hear from Bill Gates, and not me. I'm very mindful of that. Thank you for coming. Thank you for being such great students and faculty, and colleagues, and thank you for waving the Carnegie Mellon flag today.

You heard that request, and I want to emphasize it. Mr. Gates has gracefully agreed to allow you to photograph him, but flash photography only in the first few minutes. I guess photography without flash after that. It's going to be like a rock concert, but instead of lighters, there will be cell phones, put on quiet, please. But do take pictures. I encourage you to do so, because this is truly an historic day.

We have the great honor of being the last stop on Bill Gates' farewell tour. And in addition, and most importantly, you're going to be in the presence of a legend. There are not many people who have changed the world as decisively as Bill Gates has, and there are fewer still who have accomplished the kind of change he has accomplished as rapidly, and thoughtfully, and purposefully.

You know the story of Microsoft, so I'm not going to repeat it here. It's just a remarkable story. It's one of the great business stories, one of the great business successes, technology successes of all time, and a great example of technology leadership.

We at Carnegie Mellon have enjoyed a long and productive relationship with Microsoft, one that has existed for a decade. We have hundreds of alumni and former faculty who work at Microsoft. In this room today, there are 57 students to whom Microsoft has extended job offers. Now you know the real reason for Bill Gates' visit here today. I'm impressed he's come all the way out here just to recruit you all. We're also very proud that we have 12 Gates Millennium scholars who receive very generous support from the Gates Foundation for their undergraduate and graduate studies. Many faculty have connections to the company and research projects the company funds. And, of course, most specifically here at the Gates Center for Computer Science, which is rising as we speak. That was made possible by a very generous $20 million gift from the Bill and Melinda Gates Foundation in 2004, shortly after Bill Gates' last visit.

Please join me in giving a Carnegie Mellon welcome to a visionary computer scientist, a brilliant business leader, the world's leading philanthropist, and a great friend of this university, ladies and gentlemen, Bill Gates. (Cheers and applause.)

BILL GATES: Thank you. Thanks very much. I'm excited to be back here at Carnegie Mellon. The work that's going on here, as you all know very well, is fantastic world class work. As the president said, Microsoft has been very lucky to hire some brilliant people from the university, and we're continuing that tradition this year with a great new crop of people who are going to do some amazing work.

Some of our top people, including Rick Rashid, who runs Microsoft Research; Harry Shum, who has the simple job of running the development group that competes with Google Search, and many others have been at Microsoft a long time, and already made a huge difference in the work that they do. We also have a great ongoing relationship in terms of the research that's going on here, and I had fun afternoon meeting with a number of the faculty, hearing about their latest developments, talking about where things are moving rapidly, and how we do a better job in computer science of making it a tool for all the sciences; how we do a better job of using computer science to reach out to people around the world so that they get the benefits of the very rapid innovation. I know a lot of those collaborations are going to lead to great new products, and a lot of impact in terms of how technology gets used.

So I've been looking forward to coming here, and what I want to talk about mostly is where software is going, and some of the opportunities. I think people are even underestimating the impact of software, and some of the ways that it can improve a wide range of things. For me personally, this year will be a year of change. This summer I switch from being full-time at Microsoft to being part-time there, and from being part-time at the Foundation to being full-time there. And that's the first time I've really changed my work focus since I was age 17, and Paul Allen and I decided that starting Microsoft would be a great thing and, in fact, interrupted my college studies at that time. I haven't resumed them, I don't know when I might. Although with all these online courses now, you could say that I'm sort of back at college, because I'm a great consumer of online courseware. Anyway, the change will be a new thing for me, and I don't know what it will be like to have that last day, and start at new things. And so a few friends volunteered to make a little movie to help me get ready for that. So let's take a look at what they did.

We had a lot of fun making that, of course, I don't think I'll have too much spare time with the ambitious challenges of the Foundation, and I will continue to work on some of the more exciting Microsoft problems, whether it's in natural user interface that I'll be talking about, or areas around knowledge management, some fun projects I'll keep my hand in, and then the full-time work at the Foundation that is very, very ambitious. Let's talk about software. Why has it become so important, and how will that change? When Microsoft got started, software was not important. Computers were very few in number, and they were used by governments and large organizations. If anything, they were sort of against empowerment. You were worried it was going to bill you the wrong amount, and you would staple the punch card, and try and mess up the big machine. The idea that this would become the tool for creativity and collaboration, and communication, the best tool that we'd ever created, and that as we connected them together a whole new phenomena would develop around that, that wasn't obvious back in 1975.

But Paul Allen and I thought, okay, we'll do software. We'll build a platform, and encourage other people to write software. Now, there was an assumption there that we could get millions of machines out, because, after all, if you want to make it economic to spend tens of millions developing software, and sell it for $100 or so, you've really got to get that base out there. But because we made that bet, and we got that going, it became a virtuous cycle. That is, as more machines would sell, it created the market for a broader range of software, and that further drove the market for the machines, and in fact that volume allowed the price of the machine to come down. And that's why from 1975 onward, that personal computer market actually not only became significant, it actually become the center of the entire computer industry. The large machines we use today, and the big server farms, or corporate data servers, these are all based on the Windows PC architecture which, because of its volume, has come down in price, and improved in performance very, very dramatically. And so we have a large software industry.

Today, there's over a billion Windows personal computers in use. There's about two billion cell phones, 20 percent of which you could think of as a computing platform, but say in three or four years there will be about two billion of those that do have the capability to run software applications. We have software that's being used in the car. We have software that's being used in set-top boxes. So a wide variety of places where software is employed today.

Broadband connections are only in 300 million locations, 300 million people having access. So that's a smaller number, but the growth is very, very high. In fact, a few weeks ago, China passed the United States to become the largest broadband market in the world, and because China has more people, we're not likely to catch up. In fact, in a couple of years, they'll have more broadband users than we have people, so it would be very, very difficult. I think that just emphasizes that this is a very global market in terms of the innovation, the actual use of the technology, it is everywhere in the world that it's making a dramatic difference.

Now, why could we be so ambitious, and why has the personal computer taken on more and more activity? Part of it is the innovation in hardware. A key insight that got Paul and I going was that Moore's Law, the rule predicted by Gordon Moore, said that we would have double the number of transistors every two years. And that meant that we could dream about almost arbitrary levels of computing power. In fact, Moore's Law looks like it will continue out into the future more than a decade, but it won't give us the same clock speed scaling we've had in the past. So it's an interesting challenge of computer science in terms of making it easy to write programs that run across the multiple cores, because that's how we'll be using those extra transistors.

The other aspects of computing, the storage capacity, will also go up exponentially. The wireless and wired connection speeds, because of optic fiber improvements and signal processing improvements, those likewise will have the same exponential improvement. So they aren't holding us back in any way.

Some of the changes are more qualitative than quantitative. The idea that we have high resolution, so that something like Virtual Earth can display a 3D model of the world, and you can walk down the street, point at a building, go in, see what's going on. If you want to buy books, it doesn't have to be a 2D format, it can be the bookstore that was created for you with books that you might find interesting because of what your friends are reading, or interests that you've shown in the past. So making 3D reasonable, the refresh rates, the richness, we're just at that threshold.

The idea of screen technology being inexpensive, so that we can project onto the walls, and tables, and any surface that's out there in the home environments, and the office environments, there's breakthrough technology that will be in the marketplace in three or four years that will make that very possible. So whether it's your room and you want to have a theme, or the various business data in the office, all of that can be displayed.

I think one of the greatest changes that will take place is the way that we interact with software. To date that's overwhelmingly been through the keyboard and the mouse, the pointing device. And I'm not saying that the keyboard goes away, it's a very nice way when you've got your hands free, you've got the table to do it on, a nice way of particularly doing text oriented activity.

But we're going to complement it with a variety of other interaction capabilities, and I broadly talk about the use of natural user interface. Perhaps the most simple of these is just touch. You see it with the iPhone, you see it with a number of PCs out there, the idea of putting touch on the screen, and making that a way that you can interact with simple applications is very, very inexpensive. The next I'd highlight is the pen, this goes beyond touch in terms of the resolution and capabilities. Here you can take your notes in a meeting or in a class. You can look at an article, annotate it, say what friend you'd like to send that note off to, very straight-forward, natural thing. And as we get the hardware form factor, this tablet form factor that's light enough, and thin enough, and has the right price, and battery life, all the kind of reading and note taking activity will naturally move to the digital realm.

We've already seen many things that were done non-digitally that have moved to the digital realm. The print-based encyclopedia that I grew up with, the World Book my friend Warren Buffett happens to own this is basically obsolete today. He says it still smells good, and I admit neither Encarta or Wikipedia have managed to match the smell, but it's a pretty good price to get that feature. So we've had a transition, the interactivity, animation, timeline capabilities of that online encyclopedia just blow away anything that could be done in print. And all textbooks will clearly go that way. My daughter goes to a school where they've been using tablet PCs there for over a decade, and so they're one of the most advanced at how you don't need that paper textbook. And some of the benefits they've had in terms of their instructional approach, as they've learned year by year by year, are really fantastic, and not just in math and science, it's across a wide range of subjects they now have experienced teachers who've really embraced this, and now that's spreading to many other locations, and that will be heightened by the hardware and software advances. So I think the pen with ink is a very important form of interaction.

Another big one, and this goes back to some work done at Carnegie Mellon many decades ago, and continuing to improve, is the speech recognition capability. This is one of those problems where computer scientists were a little optimistic. We only missed how soon we'd have it right by a couple of decades or so. But that's okay. What it speaks to is the incredible recognition capability, and learning capability that humans have, and matching any of these human recognition systems has just made us admire how great those systems are. But with speech now in known domain, the quality has gotten quite good. So we have Tellme software, for example, that runs on the cell phones that can take a business inquiry, or a search inquiry, news inquiry, and respond to that in a very rich way. Today that software handles the vast majority of all directory assistance requests in the United States. And so we're building up a database of what we get right, what we get wrong, constantly retraining it. There will be hardware improvements in terms of multiple microphones that will help with the various noise elimination problems we have, but constant improvement now that we've hit that magic threshold. We'd say that there will be more searches done through speech than through the keyboard five years from now. And so it's one of the big bets that we're making is bringing that both to the cell phone, and to the auto, and to the living room, and to the PC environment is that speech interaction.

You saw recently Ford in all of their new models put in this thing called Sync, which is a speech interaction thing, and I've been very pleased to see the response to that. Select your music, initiate a call, even send a simple text message, that is a speech-oriented activity. And so the foundation of research over the last few decades helped, and it's been refined, and refined, and the power of the hardware finally let us say that that is moving into the mainstream.

There's another type of natural interface that I think is perhaps the most important of all, and I'll call it vision. That is the cameras that connect up to a PC are very, very inexpensive, and the software algorithms that are being worked on here, and in many other great academic centers, and Microsoft Research and others is getting better, and better, and better. So the idea of recognizing gestures, recognizing individuals, recognizing objects, it's to the point where it's very effective. And so we're putting a product called Microsoft Surface into the marketplace, it's like the table, and when you put your fingers down it sees what's going on, if you put a cell phone, it will realize that's what it is, and try and contact it with Bluetooth. If you're playing checkers, it will know, or dice, or cards, or anything that might be fun to do. It's seeing what's going on. And so instead of talking about a computer on every desk, we'll talk about a computer in every desk. In fact, every surface in your office, or your living room will be intelligent. When you go to that whiteboard and put up the schedule for a project, you can point, dive into different milestones, circle things, put up a video of somebody who you're engaged in a telepresence conversation with, all in a super natural way. And the home environment, like organizing your family photos, or picking which videos you're interested in, or playing new types of games where you have all that surface area for display, you can do something quite immersive, and quite fantastic. All of that is very achievable in the next four or five years. And so this Surface type computing, both vertical, such as whiteboard, or an intelligent mirror in your home, I'll let you think about some clever applications for that, I'm sure we won't think of them all, it's an even more impactful form of making the computer pervasive, having it be everywhere, and the software progress on that problem is the reason that I'm so optimistic that that will come into the mainstream.

Now if we think about this in terms of different activities, I'm saying that information workers, workers who work in offices will be far more empowered with information. Their ability to drill down into data, look at critical paths, understanding what's going on with their most important customers, collaborate with people at a distance, stay in touch, be alerted when something is not working well, take their key objectives and have indicators that really keep them up to date about those things. Themselves individually, program in little algorithms that relate to the domain they work in, whether it's analyzing quality, or looking at the sales process, all of these things will be available to them. And today they are only very partially empowered. A lot of their time is not spent very effectively. They don't see the external data. They don't see the patterns from that data. And so that mission of making people in the office way more effective isn't nearly done, and software is the thing that will complete that and make those jobs not only more effective, but also a lot more interesting as well.

If we look at developers, we still write code at a very low level. It hasn't changed really for 20 or 30 years. I mean, we had Pascal, and FORTRAN even before I used computers we had those things. And we haven't we're not at a level of abstraction where the data models, or the ability to make declarative statements have really gotten to be the way that, say, a business would customize applications. Big businesses today can have millions of lines of code where the differences between them could be described in terms of, say, 100 pages of English. So there's a huge explosion as you get down into executable code that leads to high cost, long lead times, difficulty of proving that it's correct, difficulty of changing. The number of lines of code really is your enemy in every way. So we need this model-based abstraction. Again, an area that a lot of good research work has been done in, but now I think we really are on the cusp of being able to say that businesses for their kinds of applications will be able to write about a tenth as much code as they do today.

Part of this is that the computing environment they run in will be far more abstract. The idea of hard disk failure, computer failure, that will be abstracted away from them, and it will be a layer below that restarting, reproviding those resources, and so their code won't have to work on those things whatsoever. In fact, for an IT person who today is managing error messages, and moving files around, or which program runs on which machine, that will be a completely automated task, both in customer data centers, and in some big pool data centers that people like Microsoft and others will provide. I often talk about these as mega data centers, because they are literally tens of millions of processors, and the whole way that you architect servers and storage, and electricity and cooling is utterly different as you move into that realm of scale. And, in fact, there are great scale economics that can work in your favor as you take advantage of those new architectural approaches. We're just scratching the surface, even the very chip designs themselves will be influenced by how they should run in this type of environment. Now, that's not to say all the computing will be there. We need a lot of computing near to the user to have these natural interfaces be incredibly responsive and effective. We'll still have computing at the corporate data center level. But this cloud computing will come and be a very important element, both the computing piece and the data storage piece.

(Continued …)

1 2 3