Harvard Conference on Internet Society 1996
Bill Gates Keynote Address
Wednesday, May 29, 1996

MR. GATES: Good morning. It's great to be back. One of the few classes I actually attended was held here, and it was Professor Finley telling us about the Odyssey and the Iliad. What we are going to talk about today is a little bit different than that, but it is an epic tale of massive changes.

The personal computer revolution has been an amazing thing. It started 20 years ago with the invention of the microprocessor, and it's really swept its way not only through businesses, but now more and more into homes as well. And it certainly has changed the way people think about creating documents. Over 90 percent of all documents are now created electronically. But in a certain sense it hasn't changed all that much. Despite the fact that you create the documents electronically, the way that you transmit them and distribute them is still overwhelmingly to print them out and deliver them, just as you would if they had been handwritten.

So we really can't say the PC up until now has been a communications tool. Now, there are a lot of reasons for that. The cost of communications has been high. And it's also very important that communications technologies are all subject to critical mass phenomenon; that is, as long as only a few people, say, have a telephone, then the telephone has no utility. It's only if you get a substantial percentage of the people that you want to stay in touch with, who understand it, own it and use it regularly, that then you get any value out of it as a communications tool.

And in the last couple of years, with several factors coming together, we're now seeing the PC emerge as the latest communications device. This all has to do with the Internet and the number of users that have gotten connected. We're now experiencing the positive spiral effect. Since we've passed critical mass, the more people that get connected, the more people want to create content and the more people are willing to send mail. And that is a powerful positive feedback phenomenon. The amount of content we have today -- it is fantastic. But it's nothing compared to what we'll have a year or two from now.

People are being drawn in. I have a lot of friends who swore they would never buy a personal computer. But if I ask them, "Okay, what are your hobbies, what are you interested in?" and then sit them down and help them browse through some of the Web pages on those topics, I can see that they're starting to break down. They're starting to think that maybe it's a relevant tool for them, even though they're not fascinated by the technology itself.

And so it's fair to say what's going on today is like the arrival of the printing press, or the telephone or the radio. And these communications tools did have pervasive effects. They made the world a smaller place. They allowed science to be done more efficiently. They allowed politics to be done a new way. They had a modest impact on how people were educated, but people were optimistic that they would make a very big change.

Now, the personal computer connected to the Internet is far more powerful in many ways than any of these other communications devices. It's not a broadcast medium where you have to have millions of people interested in something. If only a few people are interested, it works just fine. Even if they don't know each other, there's a way to connect up to each other. The marginal cost of publishing in this environment is close to zero. If you have a PC with a little bit of software and a communications line, you're as empowered as Time Warner or any other media conglomerate. Now, that doesn't mean anybody will ever visit your home page; it just means that it's there. A lot of people have very lonely home pages nowadays-- and they're just dying to have their mother or their relatives get out there and visit them.

There are really two miracle technologies that have brought about these changes. The first is the microprocessor itself. It's subject to what Gordon Moore coined and is now called Moore's law, which is that about every 18 months the power of the chip more than doubles, and that's without any increase in cost. The second is communications technology, in particular optic fiber connected up to digital switches. The rate of improvement of the speed we can get for a certain cost in switching technology is actually improving even faster than the microprocessor. And you can contrast that with other technologies, such as improvements in car fuel efficiency and work productivity.

You know, there is nothing like exponential improvement. And it makes it very hard to predict things. The easiest way to predict things that are improving that rapidly is to simply say to yourself, what if computing was free? What if you could have an infinite amount for every person? What would you do? Likewise for communications you might as well just think, what if there was infinite bandwidth connected up to every point on the globe? How would people use that? How would they reach out and think of the world differently?

Now, by some time scales this revolution is going to be very, very rapid -- something like the telephone. There was a whole generation of people who grew up thinking it was a strange device, and they didn't have to incorporate it into their lifestyle. And then the generation that came afterwards took it for granted and just never even thought about it. Computers and the Internet will be the same way. At some point they'll be so much a part of everyday life that, like piped water or electricity in your home, you'll just take them for granted. You'll think it was quaint that people once had to read the same newspaper, that it wasn't customized for individual preferences. You'll think it was quaint that if you had a curiosity about a subject you weren't able to reach out and explore that subject in a very individualized fashion.

But to make this happen will take something like 20 years. And the main reason is to get these high-speed connections to be pervasive, particularly getting them into homes around the world, will take a long time. We are in kind of a frenzy right now, almost a gold rush, where people are in some cases even going overboard, being a little bit over-optimistic about how quickly it will take place. But even though they're wrong, to have something like this happen in 20 years is really unprecedented.

Now, people often say, "Well, what was the very first day that this happened?" Well, the technology and protocols involved actually go back over 20 years ago. When I was a student here, in 1973, TCP/IP, the basic protocol that drives all this stuff, was in widespread use. In fact, it was kind of fun. There was a data computer down at Carnegie-Mellon, where they had a big disk and they'd let you just FTP [transfer] files down there and keep them. And that seemed rather fantastic, because the cost of this storage was very high. The fact that you could randomly send files down there and they would just sit there and be available, that seemed fantastic.

There are two things that have brought this into the mainstream. Certainly you can't ignore the fact that having a lot of PCs out there was part of it. But other important recent events were the arrival of hypertext capabilities, or Web capabilities, that have an ease of use that go way beyond file transfer, such as FTP, or even Gopher, which was sort of an intermediate step that allowed you to have various types of menus.

The idea of hypertext has been around since the early '60s. People like Doug Engelbart at Stanford Research, or Ted Nelson, actually talked about sophisticated forms of hypertext that still don't exist in the Internet today. In the Internet today, for example if you have an article, it's very hard to annotate that article and have people come in and see your annotations. And Ted Nelson talks a lot about that. Certainly we will have that with software over the next couple of years.

I think the key factor in all of this is the arrival of low-cost communications. To somebody who's involved in this business, none of the individual pieces are all that amazing. The way a name is resolved, the way a page comes up -- it's all pretty straightforward. But the fact that it all works at the scale that it does today; with an economic model where you don't pay any extra money for additional usage or for connections to the person next door or around the world; with traffic jams that are rather modest in nature, with technology staying ahead of that with ever-faster switches and faster optic fiber, that is an impressive thing.

It's very different from the phone system where one company had all the factors under their control and made it come together. Here you have literally hundreds of companies who provide the backbone, the software, the computers, and yet it all works extremely well.

When the PC was first getting popular it was kind of a lonely cause. There were a few people who believed in it. We'd get together and say, "Aren't we right about this? Aren't we going to overthrow those big computers? You know, we'll show everybody." We felt like people weren't paying enough attention. Well, this revolution is the opposite of that. People are almost paying too much attention. You can't get away from it. You go to watch TV, and the next thing you know there's just URLs on your TV set. And all the arcane things are there for everybody to look at. Well, I think it's fantastic, given the impact that there is all this attention.

It's the biggest gold rush that has ever taken place -- new companies, big investments. And this almost over investing is a wonderful thing for moving this system ahead very, very rapidly. The competition is quite incredible. All these protocols are being worked on in companies and committees, and so even for something like security that just six months ago was a major issue, now we have a protocol for using credit cards across the Internet. In some ways it's actually more secure than simply giving your credit card out on the phone, because the actual merchant that you buy from doesn't even get the credit card number, they just get a bunch of bits that only the bank is able to access. This is just one example of how rapidly things are advancing.

The authoring tools are also a big part of this. It's a lot easier to create pages now than ever before. It's important to remember that the machines involved here are going to continue to improve. We haven't reached any limits in terms of Moore's law or the other elements in the system. And some of these improvements change the character of what it is like to be browsing.

For example, if you have large storage capabilities on your machine, all the parts of the Internet that you are particularly interested in can be downloaded to your machine in advance. So you just set aside some of the storage and say, "I'd like to look at The Journal and Sports Zone" -- and say you have 20 or 30 sites -- by the time you walk up to your machine it will have retrieved and posted the latest updates, so you won't have any speed issue there.

With the PC of the future, you won't even have to turn it on and off. It will essentially be on all the time, so there will be no waiting. The extra speed will be used to make it more engaging, delivering better graphics. Higher-speed connectivity is fundamental here. In a sense, the speed of the network today is too slow for the mainstream. Waiting four or five seconds for a page to come up just doesn't cut it.

We will have new ways of interacting with the system, not just through the keyboard, but also with voice and handwriting -- even what I call video. With a camera that you use for video conferencing, it can watch what's going on, and it can see who's sitting down in front of the computer. You can even make gestures indicating, "Scroll that document" or "Throw that document away," and it will see that on the camera and do those things for you. So we're going to be driving the ease of use very rapidly.

One thing I think is important for video is getting very high-quality screens. The computer screen today is a terrible limitation versus reading the newspaper. The amount we can get up on the screen is far smaller. Now, the investment levels, say, in flat screen technology promise that sometime in the next decade large screens will be quite affordable. And so the surface of your desk can be entirely a flat screen. A lot of the walls in your office or your home will also be flat screens with very, very high resolution. And that greatly shifts which documents you'd like to see on paper versus on screen.

At Microsoft, we use electronic mail to send everything around. But when you get a large document, it's very typical to print it out on your local printer and then read it off of paper. Many people do this because anything more than about four or five screenfuls is just easier to read that way. But with new screen technology, that won't be necessary.

Now, when you think of the Internet as it evolves, there will actually be quite a variety of devices connected to it. There will be the device that you sit close to, which I think will continue to be called the PC. It will be so much better than today's PC that, it almost stretches the term. But you'll have both desktop and portable machines. You'll have the screen you sit further away from, which will be the successor to the TV. This will evolve by building electronics into the TV, or by using what's called the set-top box that connects the TV to the cable system, or by having a game machine with a modem that connects you to the Internet.

Today that's not happening because the communications charges are so high that pages aren't authored that way and there's some limitations. But over time, taking advantage of the TV certainly will make sense. Even the phone will be connected to the Internet. Now, it won't be much different than today's phone. It may have some voice recognition to do directory for you or the long-distance charges may be cheaper. But it will be seamless to you that it's all hooked together. You'll certainly have a small device that you can carry around and put in your pocket that will connect to a wireless digital network, and we call that the Wallet PC.

There's a lot of early, almost prototype machines in this space, but they're too expensive and they don't really connect up to the Internet. Digital wireless networks are still not pervasive today. It will be three or four years before all the pieces are in place. But being able to see a map of where you are because the device will contain a global positioning sensor, the idea of being able to receive messages, track your charges, use digital tickets and digital money, all of this will be so much more convenient with the small devices everyone will have. And there will be a companion device for the large machine. Whenever you add something to your schedule for example, it will automatically appear on both machines.

Finally, just to give you a sense of what kind of variety there will be, you will have a computer in your car that once again uses a map, shows you traffic conditions, lets you find the nearest restaurant, whatever your particular tastes are. And this will be commonplace, with all of these machines hooked up and working together.

There are two main things that really hold back the Internet. Why doesn't it happen overnight? Well, first is just the rate at which people can absorb new technology. How quickly are people drawn in to this? How quickly does the content get up there? That just takes time. The fastest that any new communications technology has been broadly adapted so far has been about 20 years. Now, if it was just the suffusion rate holding back, it'd probably be less than a decade.

The other item, though, is the speed of these networks. Unlike PC performance or storage which improve exponentially, communications networks, especially when it comes to the local part of the connection, are much harder to improve because there is a fixed cost of going in and putting in new wires. So there are three different generations of communications bandwidth. First, there's today's phone system, which is called narrow-band, and uses a modem that connects to 28.8 bits per second. In that world, text comes up super-fast, but images take maybe five or 10 seconds to appear, so it is quite limiting.

Then there is the next generation that's called mid-band.. There are many examples of this -- ISDN, ADSL, PC cable modems. We're just at the very start of this phase. They are all fairly expensive today and they're not all hooked up. And in the case of things like the cable system, there's a pretty significant rebuild that has to be done on parts of the system before it has two-way capability. It certainly will happen, but by the year 2000, of all the homes in the U.S., at most 3 or 4 million will have these mid-band connections. And so most, if they're connected to the Internet, will still be narrow-band and we will still be trying anything we can to deal with the bottleneck that this lower speed creates.

With networks that are mid-band, the still images are very, very fast and you can start to do fairly low-quality video, probably good enough for a video conference but not good enough that you'd watch a two-hour movie through that kind of resolution. The holy grail is broad-band. That's where you get data rates over a couple of megabits. And the beauty of that is you can start to do very high-quality video. So as you track how this spreads out, it's very important to watch communications speeds. Those are the key holdbacks.

The U.S. is in the best shape of any country, because now with our telecommunications deregulation bill, we have a very strong cable industry that has an incentive to compete with the phone companies for the most difficult piece of the business, local access. But having two companies in competition is not as good as having 10 or 20. And assuming there's no radical breakthrough in wireless technology, with the kind of data rates we have now, the only way for urban areas to become connected is through wires running into the home.

Now, as I said, there's a lot of competition, and that leads to great industry issues. There are many different operating systems. You heard from what I might call some of the minority representatives in that area yesterday. And it's fine, they're very articulate. They do wonderful work. One of them is even quite humorous. And, you know, we'll keep him around as long as his jokes are good.

The world of operating systems becomes more and more homogeneous over time. Today something like 85 percent of the computers on the planet run the same operating system. There's sort of a positive feedback cycle here -- if you get more applications, it gets more popular; if it gets more popular, it gets more applications.

Now, historically PCs were very low-end computers. And it's only because of Moore's law that they're now moving up into the more demanding tasks: scientific visualization, business data bases and so on. There are still some pieces that are being put into place to make that all happen. And the competition is very healthy. The only thing to take away from the industry issues is that we're going to keep each other honest, and no one's going to get something that's completely proprietary.

There will be a variety of machines that hook up to the Internet. And the software that does all this is going to be fantastically inexpensive. In fact, it's Microsoft's position that this software will just be built into every operating system. So, if you get Windows, you'll get your browser and everything you need. If you get a Mac, you'll get your browser and everything you need. You won't even have to buy add-on pieces. The history of computing is like that. Things that you first had to get separately, are increasingly built into the system.

And so you have this dynamic of deciding when you use a PC versus a more expensive machine. And more and more that moves to the PC. There is at least a paper battle about a PC versus something less, something whose virtue is its incompatibility or that somehow something has been taken away. This is the so-called network computer. And, you know, it's a great debate that will be raging for a long time. There's going to be a key milestone, though, which is when they actually make one. And at that point you'll be able to look at it and say, "Well, what did they take away? What thing is left?"

I think PCs will get less expensive. They've got to get less expensive. They've got to get down to even $500 for all of this to be pervasive. Absolutely, that can be done. The market's always making a tradeoff, when you have so much innovation, between using the innovation just to get a more powerful machine at the same price or the same machine you've had now at a lower price. And many people have offered inexpensive PCs. They haven't sold that well because the market to date has opted for more power at the same price.

But as the absolute level of power gets to a certain level, as we get the ability to do advanced 3-D graphics, motion video and all of that, then eventually the marketplace will move a lot of that innovation into lower prices. In fact, just recently the price of memory has now come down a lot, which is a very positive thing. So certainly there'll be a variety of devices. What's a PC and what's not will be a tough call there.

When we think of the Internet, there are so many different scenarios that are important to consider. Electronic mail -- it's not just messages, it's also a replacement for paper forms. Think about bills today and how much overhead there is when you mail out a bill. You've got to print it, and then mail it. Somebody's got to look at it, figure out if it's wrong and they call somebody. They mail something back in that then goes into a system. Compare that to simply having an electronic bill arrive in your mailbox. If you don't think it's right, you send back a mail message saying, "Hey, why is this this way?" You get a quick response to that. The overhead you can take out and the convenience you can bring to this is very dramatic.

Information publishing about products -- companies like to get the data out there. Clients just like to share information. Every night the new DNA sequencing information goes up onto the Net, and the biotech companies go out and look and see if anything that fits the patterns they're trying to find is there. So it's advancing the progress in science. Conferencing-- we'll have a lot of software to allow for conferencing across the Internet. If you want to negotiate a contract, you'll be able to both talk and edit the contract at the same time. If you're confused about what's going on with your PC, you won't just have the phone connection, but you'll be able to share what's on your screen and have somebody step you through exactly how to do something. So support will be greatly improved.

The very mechanism of capitalism is greatly improved by the electronic market. Some large markets like currency are well-mediated today. But most markets, like finding a good lawyer, a doctor, a consultant with particular capabilities, or even a babysitter, are not well-mediated markets in the sense that matching all the buyers and sellers with the right characteristics is hard to do. Well, electronically, that's immensely easy to do. So you'll see things like classified advertising and the Yellow Pages move over to electronic form as the system becomes more and more pervasive.

One of the key areas that people haven't figured out yet is how does this impact learning? Clearly there is a lot of potential here, but universities really haven't decided how to redesign themselves when all of this expertise is out there on the Net and easy to browse. What does that mean for how they function?

One of the flavors of the Internet that will catch on very quickly is the Intranet. That's where you share things internally. You could do this with PCs networked together previously, but knowing server and file names and when the file was last updated was just too much work. So now you have these pages that are completely self-explanatory, and if you want to go do something, you just click on a link to do it. People are now sharing documents this way. And it just uses today's software, perhaps new versions of it at very low incremental costs, because the PC and the network are already in place.

One of the applications is to reduce the cost of PC ownership, using the Intranet to transmit new software updates and any information that people want. So when you go into companies now, you'll see that all the human resource data, all the financial data, sales tracking information and all the project status information is on these pages that people just navigate around. It's an incredible thing. I doubt there's a company represented here who couldn't go further in using this type of sharing and getting more value out of the PC investments that have been made.

There are a lot of software pieces that go into this -- including enhancing the operating system and having server software that makes sure the data is dished up very quickly, even from structured databases that have very good protection and clear capability. You'll have actual code in these applications -- Basic code, C++ code, and Java code. And you'll have the normal sort of text and images. All those things have to be brought together to get very rich information sharing.

One of the things that's hard today is navigation. The way you move around the Internet is very different than the way you move around your local storage files. So we're going to unify that so you use the good things about the Web when you're working with your local information. Then you only need to learn one metaphor. Instead of files and pages, you have it brought together so every time you look at a directory, it's like a Web page. You can think of a directory today as a degenerate Web page where all you have is the link, that is the individual file names for each link can take you to that file. Here you can have lots of explanatory text, and it becomes transparent to you, except for the speed, whether something's local or remote.

We also need to take these pages and make them active, so that you can get sound, or a movie, or a ticker with news and stock data. That's starting to be a big thing. Now, we need a little technology to do this, and this is an industry thing where we're pushing an approach called Active Controls and Active Documents in order to achieve this.

So the quality is going to be there. Don't let anybody confuse you or question if security will always be broken or traffic will be overwhelming. The industry will solve those issues. The broader challenges and things that the industry can't address alone are some of the things I have here. The issue of privacy. People probably don't realize today how much information about them is stored electronically. It's pretty phenomenal. And with these tools, you can gather the little bits of information together and create what most people would view as an overly invasive profile of somebody's activities.

So the question is, if you buy a product from a company, what right do they have to take your name and give it to other people and make that information public? This is going to be hotly debated. And it has nothing to do with technology. I mean, once the rules are known, the computers can enforce them. But it is a deep problem. People who think that politically we should go to the extreme and have ultra-privacy, that has some real drawbacks because then you're not able to use the information on behalf of that person and let them know about sales or guide their navigation in a richer fashion.

One of the bigger issues is the haves versus the have-nots. Now, you only get this issue if you actually believe that this is great technology. If you think it's all just a bunch of computer guys goofing around, then don't waste any money on worrying about pervasive availability. On the other hand, if you think this is like books, where everybody as part of their learning experience or throughout their life should have access, then we have to figure this out, how do we use schools? How do we use libraries? How do we allocate resources to make these things available? There's a lot of great pilot programs in these areas, but it's a tough problem. And the issue of the have versus the have-not has many dimensions. It's got rich versus poor, urban versus rural, young versus old, and perhaps most dramatically, developing countries versus developed countries.

We have a lot of security technology in the network. I said that's going to work very well, but from time to time that security technology will break down. And so having ways that you can fix that and bound the difficulties that this creates is very, very important. Recently in Japan they had electronic cards for playing pachinko, and somebody broke the code on those cards and made all these fake pachinko cards. So a couple of large companies in Japan lost $100 million because of all that essentially counterfeit currency that got out there for playing pachinko.

There's a huge political issue with security, which is that the U.S. government makes it impossible for software companies here to export decent security technology. That's because the spy agencies still want to be able to tap into mail that's sent around the world by different people. And unfortunately for them, this technology is fairly pervasive, so there's this debate whether they should let software companies, like Microsoft and others, actually have decent security.

Another scheme they have is for all the security you give the keys to the government. So, if the government wants to look at what you're doing, they can do that. Well, the problem is that the people who the government might want to look at, they are the ones who won't turn the keys over. So this whole key escrow debate is a big issue that has a political element to it.

A final issue centers on what should be allowed on the Internet. There are some idealists who think the Internet should be a lawless environment. That is, that you can libel somebody, you can promote securities that are fraudulent, you can steal software, you can do anything out on the Internet. It seems kind of neat to think that, "Oh, this is something that no government will touch." Well, that's not likely to take place. The Federal Trade Commission has started to look into scams here. The SEC is starting to look into problems that have come up here. Certainly companies that sell copyrighted software or people with intellectual property are very interested in rules being enforced.

There are some tricky issues, because the fact that the network is global means that a national government has a hard time reaching out to various Internet sites. And telling all the communications providers in a country to block an Internet site is very difficult, because it can move very quickly. So censorship becomes very difficult, and yet there are going to be some rules about what's out there on the network.

So my belief is that there are some very deep implications. It's great to see early debate about what all this means. The way we learn, the way we elect politicians, the way we spend money, even the way we entertain ourselves, is destined to change, and on the whole in a very, very positive way.

Thank you.