Dr. Cerf is widely known as the father of the Internet. He is the co-designer of the TCP/IP protocol, which is the computer language that gave birth to the Internet and which is commonly used today. In December 1997, President Clinton presented the US National Medal of Technology to Cerf and his partner, Robert Kahn, for founding and developing the Internet.

In addition to his packet-switching activities, Dr. Cerf enjoys, is knowledgeable about and, within appropriate limits, deeply appreciative of fine wine. Dr. Cerf.

Vinton Cerf: I can attest, by the way, on that last point, that there are some places in Ann Arbor that have some very fine wines to offer. We found some at Escoffier last night, hiding in the cellar, and we liberated them to good advantage.

It is indeed a pleasure to be here. I'm sorry; you're pointing at me to go and push the button, is that right? Thank you. (Looks at the screen, which displays his Power Point presentation.) All this technology - well, that's not it. It says, "No input." There will be a brief pause while we convince the system - there we go. (Laughter.) Except now I can't see it. There we are. How's that?

Anyway. It's a real pleasure to be here. I have come to Ann Arbor now and then, largely on missions associated with reviewing the programs that take place here, the research programs in Ann Arbor, and I've made many friends, including Doug Van Houweling, Dan Atkins and others. It is always a great joy to return to see old friends and to make a good many new ones.

My task is complicated today. I'm going to try to characterize what has happened, what is happening and what will happen on the Internet in about 20 minutes' time. So let me start. First of all, will you kindly set your modems to 50 gigabits per second and we will proceed.

In simple terms, the Internet is simply the largest network of networks ever assembled. It has well over 200,000 networks around the world, connected together in this system, using the TCP/IP protocols. The Internet protocols, the IP protocols, were designed to operate on any communication substrate, and that's why I have a t-shirt which says, "IP on everything." (Laughter.) No, no; Internet protocol. I even got one for my dog, so I could take her along on walks.

It started out as a military research project, trying to understand whether packet-switching as a technology, which Philip mentioned earlier, would be useful for implementing computer-based command and control in the military. It quickly emerged as a very successful, new way of allowing computers to share communications resources and to exchange information. We explored this not only using telephone circuits as a way of connecting things together, but mobile radio and satellite as well. And that led to an interesting problem, because we applied packet-switching to different kinds of communication networks, implemented it slightly differently in each case, and then had the problem, how do we connect these different kinds of nets together? That was the "inter-net" problem. And that problem was tackled by Bob Kahn and me in 1973, and what emerged out of that was the design of the Internet.

These dates should suggest to you that it's been a long time in coming. The National Science Foundation Network was one of the key developments that brought the academic and research community together, and it's quite fitting that we should be talking about that because, here at the University of Michigan, MERIT - an organization devoted to computer communication - was instrumental in the implementation and operation of the National Science Foundation Network, which was retired not very long ago, in 1995.

To give you some sense of what has happened over the last 20-some odd years, there are a few key dates that I show in this slide; I'm not going to go through every one of them. But what I do want to stress is that Internet has only in the recent past become a commercial activity. Internet has for most of its life been part of the research and education environment. It wasn't until 1986 that the first commercial product - namely, routers, the things that actually move packets of information through the network - became commercial products. Up until that time, we used to wrap graduate students around computers and try to get keep them running that way. But we ran out of graduate students, and so eventually somebody like Cisco Systems recognized that they might be able to sell these things.

It wasn't until 1989, just nine years ago, that any commercial interconnection was permitted on the Internet. Just nine years ago, I got permission from the federal government to connect MCIMail, one of the first commercial electronic mail services, to the Internet, as an experiment. And shortly after we did that interconnection, of course, everyone who had a commercial e-mail system insisted on having the ability to connect to the Internet as well. And of course the experiment never stopped.

In 1990, the first of the packet-switching networks was retired, the ARPANet. And the first commercial services emerged on the network; only eight years ago, the first commercial Internet services came along. And it wasn't until 1994 that the public became aware of the Internet, primarily as a consequence of the commercializing of the World Wide Web, principally through Netscape Communications. And so that's only four years ago, and NSFNet was retired only in April 1995, just three years ago.

So, much has happened, most of it just in the last few years. So we think we are still in the middle of what I would call an Internet gold rush. It's important for you to remember something about gold rushes. The people that make money in the gold rushes are not necessarily the people who are looking for the gold. The people that make money in the gold rushes are the ones selling the picks and shovels and tents to the other people who are looking for the gold. And that's what the telecommunications industry is doing with the Internet; they are selling the electronic equivalents of picks and shovels to other people looking for gold in the Internet.

If indeed we're in the middle of a gold rush, then some of these statistics should be indicative:

• There are 2.1 million domain names registered in the Net - you know, things that read www.mci.com. They are being registered at the rate of 150,000 a month. In fact, some companies are so eager to register their trademarks as domain names that we get such amazing things as www.pimples.com and www.diarrhea.com, both of which are trademarks of Procter and Gamble. I haven't signed up for this yet, but I've always wanted an e-mail address: vint@diarrhea.com. The only trouble is, no one will be able to get there because they can't spell it, but . . . different problem.
• There are estimated to be 30 million or so computers in the Internet as of January this year, and I need to remind you of something. Thirty million computers is a lot of computers. On the other hand, compared to the telephone system, which has 828 million terminations - 128 million cell phones and 700 million wire lines - the Internet is pretty small by comparison, despite all the hype.
• On the other hand, the Internet is growing much faster than the telephone network is. It operates in 240 countries today, not all of them uniformly penetrated. China is on the Net, but not nearly in as penetrated a way as the US. Interestingly enough, the United States is not where most of the highest per-capita instance of hosts is. Finland is the country that has the largest per-capita number of computers on the Internet, and all we can figure out is those long winter nights in Finland must leave you with nothing else to do but surf the Net.
• No one knows for sure how many people there are on the Net, but it's estimated that there were 70 million on them, perhaps about half in the US, as of the beginning of this year.
• If you look at the trend in domain-name registration, even if it just continues at the current rate, there will be 7 million domain names in the Net by 2001. And as far as the number of computers in the Net goes, the estimates range somewhere from 100 million to 200 million in the Net by the middle of 2001. That's important, because that's starting to get to the same size or order of magnitude as the telephone system. And that means that it is an important thing for the telecommunications companies to pay attention to, because of the scale.

Just a few other observations. In the MCI backbone -which, by the way, has not actually been sold yet. The deal that was announced yesterday won't be consummated unless the merger of MCI and Worldcom happens. If that doesn't happen, then the sale won't go through. So - just a small little detail - in that backbone, we carry 208 terabytes of traffic a week. That's over 800 terabytes a month. I don't even know what to suggest to understand that figure; it's just huge. The only people that really seem to understand it implicitly are the people in Congress who deal with numbers of that scale all the time. Seventy-five percent of that traffic is World Wide Web traffic, and that truly is an amazing statistic, given that that was zero about five years ago. There wasn't really any significant World Wide Web traffic in 1993.

There are estimated to be a million and a half Web sites and 350 million Web pages. Now, some of you I'm sure are familiar with the theory of "dark matter" in the universe, the thing that holds the universe together even though you can't see it. Well I think that there's a lot of "dark information" in the Internet that is not visible to the World Wide Web, because it's not expressed in the language that the Web search engines can use. Instead, it's all locked up in big data bases. So in fact the 350 million Web pages may be a small fraction of the total information in the network if you take into account these gigantic databases that can only be accessed by going to a Web page and then formulating a query to pull the information out.

Depending on whose network you look at, traffic growth can be anywhere from 100 to 1,000 percent a year. And clearly, there continues to be very low barrier to entry into the Internet service business, because there are 7,500 Internet service providers around the world, 4,500 in the US. And the projections are that there will be somewhere between 300 million and a billion users by the end of 2000. I can't believe the billion number, to be honest with you, but 300 million users I think I could imagine in that amount of time.

I often worry that the rate of traffic growth in the Net will in fact consume all of the available fiber capacity in our network, and in fact we projected that by the third quarter of 2001, we will need to supply as much capacity for the Internet as we do for our voice system. And so, in a panic, I double-checked to find out whether we were going to run out of fiber. And it turns out that we are far from running out of that kind of capacity. Typical optical fibers today in our network carry anywhere from 10 to 40 gigabits a second of transmission speeds, and in fact those same fibers can carry trillions of bits per second. You simply have to put more, different colors of light down the same fiber; then the photons don't interfere with each other, as long as they're different colors. So we are far from consuming all of the capacity of the network; we just have to light more colors through each fiber.

The big problem is the routers, the things that switch packets through the network. They don't run as fast as the fibers can actually deliver the bits, and as a result, every morning I get up and say a little prayer for Cisco Systems and 3Com and Bay Networks and some of the new startups, say Juniper, hoping that they will produce what I've been calling a BFR, which stands for Big Router. (Laughter.) You're looking at a man who doesn't just believe in miracles but he relies on them, so the BFR has to show up sometime this year so I can keep up with the traffic demand.

There's another interesting phenomenon in the Internet, and that is the continued and very rapid growth of internal networks, corporate networks, that are used to implement and automate business processes inside each company. The estimates by Zona Research are that there will be something like $28 billion worth of business a year in internal or intranetwork services, and $14 billion a year in the public Internet, for a $42-billion-a-year industry by the end of 1999. What's interesting about this statistic is that it probably portends a slingshot effect, because after you see all those internal networks constructed, companies will discover that they can't take advantage of what they have implemented fully without exchanging information with other companies. So having made this big investment in internal networks, there will be a slingshot explosion of demand for public Internet service to allow inter-corporate exchanges, and we'll see a little more about that as we go on.

Many of you are already users of Internet services, so I'm not going to bore you by going through a list of all of them that are available. But I do want to let you in on a little secret. Many companies have tried to make money out of video conferencing. By and large, it has not been a terribly successful business. And I suspect that successful and widespread video conferencing is going to come from an unexpected quarter. And I'm thinking down here of interactive, multi-user games. As you know, video-game manufacturers have penetrated some 90 million households in the United States. The age group that plays those games would dearly love to play them in groups scattered around, not necessarily all in the same physical location. So the video-game manufacturers are building Internet chips into the video games so that they can be plugged into the Internet and the kids can play games in a distributed fashion. Well, if you know this age group, you can imagine they would love to hear each other screaming and yelling while they're playing these games, so putting a microphone on the video game shouldn't be a big problem. And if they could see each other, especially when you shoot somebody out of the sky and you want to see their face as they head straight for the deck, putting a hundred-dollar video camera on these games would also not be a technically hard thing to do. So, now, let's see what we've got here. We have a little device and a collection of people scattered around, distributed. They can hear each other, they can see each other and they can shoot at each other. It sounds like a video conference to me. Moreover, the people who are making these things are making them as consumer products, so they're going to drive the prices down to something that consumers can buy. So for all of you out there who were planning a video-conference business, take note. It may very well be that the video-game manufacturers are the ones who will be making the real money.

Around 1995, "electronic commerce" became a hot buzzword in the Internet, and there were lots and lots of experiments by companies trying out ideas for making money in the Net. And I love the names of some of the companies. Cybercash. Digicash. First Virtual Holdings. First Bank of Internet. We had one called Marketplace MCI, which we eventually decided was not a money-maker and we stopped offering the service. I don't have time to explain to you why it was that we didn't make a success out of that business, but we learned some important lessons, not the least of which is that catalog sales can't be mirrored directly on the Net as Web pages, because it takes a long time to put a fancy color display up on a 28 kilobit modem, and we didn't pay attention to that in the early stages.

In any case, one of the things we did learn is that encryption technology, cryptography, is very important to electronic commerce. It's needed to protect the confidentiality of transactions. It can be used to authenticate the source or the parties in the transactions, and it's our view that high-quality cryptography needs to be available essentially worldwide so that electronic commerce can be protected worldwide. Unfortunately, that view is not shared entirely by the administration here and in fact in other parts of the world, and so there's been a tension between the commercial sector and the government sector as to the rules for export of cryptography. My view right now is that the attempt to control export is going to come to little for the simple reason that rather than exporting the technology, companies who need it will simply implement it in different countries, as required and in accordance with the standards. There won't be any export, but the separately implemented systems will interwork because they are standardized and the export won't be the method of controlling them at all. So in the end, there will be good quality crypto everywhere as needed.

In fact, let me come back to a little bit on electronic commerce. We are seeing the beginnings of some very serious efforts to create electronic commerce environments that allow multiple companies to interact with each other. Amazingly, and appropriately here in Michigan, General Motors, Chrysler and Ford are cooperating with each other to develop standards to allow them to interact with their suppliers, which number in the thousands. The idea is to be able to place orders, confirm orders, send specifications, negotiate engineering designs and the like with those suppliers in a confidential way, which means that this system has to have good quality confidentiality using cryptography, among other things.
Bellcore has been selected by the three companies as part of this automotive network exchange effort to do the architecture of the system. MCI among others are supplying the communications capability in the Internet for this experiment. The entire outcome, I think, will light the way towards electronic commerce in other industries, and so it's of great interest to all of us to see how that goes.

In terms of transaction value on the Net, we are anticipating, at least Forester Research is anticipating, that there may be as much as $327 billion worth of value exchanged on the Internet during the year 2002. My first reaction was, "That's a big number," until an economist friend of mine pointed out to me that the gross world product in 2002 would be $30 trillion, which mean $327 billion is only 1 percent. And so, having lived in Washington for 22 years, Mr. Ambassador, and having not yet caught Potomac Fever, I can stand up in front of you with a straight face and say that $327 billion is a small number. In fact, it may even be a low estimate of the value of transactions that take place on the Internet in four years' time.

You can imagine, with those sorts of numbers floating around, there are a lot of people wondering whether they could tax any of those transactions to raise revenues. Just to give you some vignettes of what's happening in the commercial sector, Dell Computer reported that it sold $6 million a day worth of computers on the Internet through its Web pages in October and November last year. Their more recent report said they were selling $5 million a day, $4 million in the US and a million in Europe, and I guess the reason that it came down a little is that in October, November everybody's buying things for Christmas, and that probably represents a little uptick on a seasonal basis.

What we learn from watching Dell is two things: First, Dell has relearned and reapplied what Sears Roebuck learned many years ago, that by using catalog sales they were able to aggregate a very distributed market into a common central market by simply sending the catalogs out and getting orders back. Not counting the value of the catalog for other purposes. "Dear sir, I'm sitting in the smallest room of my house. I have your catalog before me. Soon it will be behind me. Sincerely yours." Of course that same thing doesn't work in the paperless world. However, the Internet does allow you to aggregate widely distributed markets that might be impossible to reach or impossible to even create in the real world, because the markets are just simply physically too distributed to treat with a single physical facility.

The other thing that's interesting is that Dell's success indicates that products that require configuration are very successful in the marketplace, and certainly computers, of all things, do require lots of choices, options and configurations.

There are sites like the Yahoo! search site that are beginning to generate advertising revenues, which continues to be a big speculation in the Internet. How many sites can actually be supported by that? I don't know the answer, but we are watching with some interest the successful ones.

We are seeing niche and specialty stores being very successful on the Net. Looking for rare wines? Rare books? Old automobile parts? These turn out to be very successful businesses on the Net. And of course everybody's favorite example of a bookstore, Amazon.com, recently reported turning a $66 million quarter, which is a run rate of over a quarter of a billion dollars a year, in less than a year. It's the fastest-growing company, I think, in the history of the Net.

Another thing that we're seeing is that the Net is being used to supply self-service to customers. Whoever invented self-service deserves the Nobel Prize. I mean, think about this for a moment. The customer serves herself, so she can't complain about the quality of the service. Or if she does, she has herself to blame, and the owner has less expense because the customer is taking care of this service. The Internet is beginning to be a tool for that same thing. Instead of listening to music on an 800 number call, customers go to the World Wide Web and interact and get service on a personal and a much more rapid basis. Companies put up frequently-asked questions and their answers. Some companies have actually put up forums for customers to talk to each other about problems they're having. Of course, this can be dangerous, because sometimes the customers get together and gang up on the company. You have to think twice about whether you're going to create a forum like that.

MCI implemented an interesting feature. We sell 800-number services that people use to call on the phone, but we also implemented a version that works on the World Wide Web. You click on an icon on your screen and make a telephone call through the Internet to an 800-number call center, which goes out through public-switch telephone net through a gateway. And the way this works is that the call-center operator receives the call and also gets information about where the customer is on the World Wide Web, so both the operator and the customer can see the same thing. The operator actually can cause pages to come up on the customer's personal computer, and as a result - if the call, for example, was about a billing problem - both the customer and the operator could see the same thing at the same time. We even went to the trouble of putting a small camera in front of the computer-center operator, or the call-center operator, so the customer could see the call-center operator when the call was placed, but the video wasn't very good, so we didn't make that a standard part of the product.

What I want to alert you to is that all of this roiling inventedness on the network, all these experiments in the commerce world and everything else are raising a host of policy issues, many of which will be addressed in this symposium, and I'm delighted to see that, because they need the attention and scrutiny the symposium will give them. Taxation, for example, has come up as an issue in local, state and national forms. Many legislators look at the Internet as a potential source of revenue. And although the federal government has, I think, rightfully said that we should go slowly in the imposition of any taxes on the Internet merely as a consequence of having carried out a transaction on the network, it's fair to observe that if indeed the value of transactions on the Net should skyrocket in the way that is projected, eventually we will have to figure out how to supply tax revenues through this system. The problem is the Internet is global in scope, and even in the United States alone there are 30,000 taxing authorities. Can you imagine if every single one of them decided to tax every Internet transaction, how complex this would be for every Internet service provider. "Let's see, that book will be $19.95, and in a couple of weeks we'll let you know what the tax is as soon as we figure it out."

Moreover, there are some serious problems with copyright, because information in digital form is readily replicated and distributed. That is an issue, I think, that is going to receive continuing consideration. Privacy, confidentiality, integrity of transactions and authenticity are similarly of great concern in this system. The actual governance of the Internet itself, the way in which its address space is allocated and domain names are assigned, have become major issues at the national level in many countries, including our own.

On the legal front, there are all kinds of open questions about the legal framework in which transactions take place on the Net. Who is liable? Does a digital signature mean anything? What happens if I engage in a transaction on the network and I have a dispute -- the supplier failed to supply, the customer failed to pay? How should I resolve the dispute? Nobody has good answers to those questions. As we begin to put telephony and video and radio on the Internet, should they be regulated in the way that conventional services are, or should they be treated as a wholly new kind of service? Of course you can imagine that MCI would prefer that it not be regulated at all. But those are open policy issues that need attention.

The Internet itself is beginning to carry many of the conventional media within its services - video and telephony and radio. Radio is working very well. There are about 3,000 radio stations that put their audio up on the Internet. Some of them don't even bother with the radio transmitters anymore; they just put the audio up on the net. It's a very peculiar feeling, though, to think that ordinary radio has a limited audience, defined by the power of the transmitter, whereas on the Internet, anybody who can get on the net can tune in on that radio transmission. I don't know what that does to the economics of radio, but it certainly changes the audience structure, and therein will lie some interesting experiences as time goes on.

Video is less successful so far on the Net. If you've ever looked at an Internet video, it shows up in about a two-inch screen, maybe six or seven frames a second if you're really lucky. It looks like a 1910 movie; I always expect to see a little subtitle showing up on my laptop saying, "Congratulations. You've turned your $5,000 laptop into 1928 television set." However, we will get there. It's a matter of capacity of the network and processing power in these devices. Already there services deliver good-quality, 30-frame-per-second video, at 400 kilobits a second. Delivering that to the customer is still an open, big challenge, because the last mile is still one of the big bottlenecks in the network.

In any event, we are being expected, as Internet service providers, to support these new kinds of applications in the network. And it's requiring us to re-engineer the insides of the Net to handle the quality of service that's required.

It does seem, as time goes on, that more and more things are going to be connected to the Internet. In fact, it's my expectation that we will see a lot of appliances that we don't normally think of as Internet devices on the Net. Already we're seeing Web television; we're seeing cellular phones that have Internet capability in them. Let me ask you to imagine what happens if your videocassette recorder is hooked up to the Internet. This could actually be a good thing. You can imagine pulling up a Web page and clicking on the television programs you want to record, and instead of grabbing your 11-year-old to go program the VCR, you can send the instructions through the Internet to the VCR and get it to do what it's supposed to do. But there's something even more important. Once the VCR is on the Internet, it can find out what time it is and get rid of that damn flashing 12:00 that's on everybody's VCR.

Let me make one little leap forward for a moment and suggest to you that at the edge of the 21st century, and as we enter into this new millennium, one of the other things that's going to happen is an overwhelmingly successful collection of bioelectronics devices and applications. I think we are entering a period when infirmity may become an affectation and not an affliction. To give you a concrete example of that, let me describe for you what happened to my wife just two years ago. She was deaf for 50 years. She lost her hearing when she was three years old, due to spinal meningitis. She wore hearing aids, but she had a 95 decibel loss, which for all practical purposes means she was profoundly deaf. Two years ago, she got on the Net, learned about cochlear implants, and had an implant done at Johns Hopkins University. It was a remarkable experience. This was an outpatient operation. She went in in the morning, had the operation done, the implant was inserted into the cochlea, and she came home that afternoon. Now, they hadn't turned it on yet. It took a couple of weeks for her to, for all of this stuff to heal. Then she went back to Johns Hopkins to be "activated" - sounds vaguely religious. I didn't go with her. I was occupied with something I couldn't avoid, so she was up at Johns Hopkins, they turned this thing on and programmed it. She wears a little computer that's about the size of big pager that takes all the sound in and figures out how it is going to stimulate the auditory nerve directly in order to simulate what the cochlea normally does.

Twenty minutes after they turned this thing on, she called me on the telephone. We hadn't had a telephone conversation in our 30 years of marriage. You can just imagine what that was like. Well, we can see, now I have a terrible problem; I have a 50-year-old teenager in the house. (Laughter.) I can't get her off the telephone. She even goes around listening to pre-recorded books with a Sony Walkman. It's truly amazing. This little gadget can, with patch cords, can be plugged into any audio device, right? So when she's on an airplane, she plugs right into the arm rest, you know, so she doesn't have to wait for the headphones. It's truly an amazing experience.

What is going on, though, is that we are understanding more deeply now how our bodies work, in particular how our neural systems work. Carver Mead, who was at Cal-Tech, told me not too long ago of his work on an artificial retina. It's probably in the same stage of development as cochlear implants were 15 or 20 years ago. And one can imagine, if we know how to signal correctly, to signal neurals correctly, to stimulate them, as we do in the cochlear implant, that some day even spinal cord injuries might be bridged by suitable electronics that can patch across the break. And that, I think, is something truly to look forward to.

Let me turn briefly to information infrastructure, because that is indeed the heart of today's and tomorrow's discussion. An information society is what we think of today as something new, but it has always existed even since the first cuneiform clay tablets; it's just that the technology is different. What is basically different is that the information is now software-enabled. Our software is facilitating communications. It's allowing people to share pools of information in ways they never could have shared before. It allows collaboratories to be built, where people who are physically distributed can cooperate with each other. We're going to hear about successful collaboratories as soon as they can get me off this here podium.

And finally, I have to tell you that the sky is no longer the limit. It took 25 years for the Internet to get where it is today, at least from my perspective. And if we look 25 years ahead, I fully expect to see colonies on the moon, on Mars, and maybe even on some of the moons of Jupiter. And it is time now to start designing the interplanetary Internet. And I can tell you, it's already begun. I've started work with the Jet Propulsion Laboratory on a design for such a system, and we hope that we will get our work done in time for those colonies to show up in 25 years time and be on the interplanetary Internet.

Finally, one last prediction. I talked yesterday to a man who is building Internet chips. He's building the protocols on a little integrated circuit. He says he can do it for 35 cents today, which says in 10 years' time or even less, he'll be able to do it for very, very little money, which suggests to me that by 2005, the light sockets will be on the Internet, and when we're controlling the lights, we won't be turning the current off and on. We'll be sending SMNP Internet control packets to the light sockets, telling them to turn the light off and on and get brighter or dim.

That concludes my formal remarks to this august body. I am looking forward to hearing some of the examples of this technology that are already in place and light the way to the future. Thank you very much.