Jessica Rosenworcel talks with Vint Cerf
If you've ever, I don't know, sent an email or used the internet, you really should know this episode’s guest. Vint Cerf, one of the “Fathers of the Internet”, co-designed the TCP/IP protocols and the architecture of the Internet. He also helped to develop MCI Mail—the first commercial email system to be connected to the Internet. Hear about his journey to the development of these world-alerting protocols, what he sees as the future of Interstellar communication, and why he always wears a three-piece suit when hoodies are standard in the tech sector. Conversation recorded in April 2024.
JR: Welcome to the FCC, and welcome to First Conversations, which is our podcast slash speaker series that puts a spotlight on the barrier breakers, glass ceiling smashers, and innovators who have helped to shape modern life. Each of our guests is a trailblazer who cleared a path for others, and you get to hear a little bit about what it took to get there. Now, as most of you know, I'm Jessica Rosenworcel, the Chairwoman of the Federal Communications Commission.
VC: Yay—first woman!
JR: It's already starting very well. Alright. So, if you've ever, I don't know, sent an email or used the internet, you really should know today's guest. I'm talking about Vint Cerf, because along with Bob Kahn, Vint developed the protocols that allow computers to talk to one another. And for this breakthrough, he is righteously recognized as one of the Fathers of the Internet. But of course, this is just one of many titles that Vint Cerf can claim. He's also a recipient of the Presidential Medal of Freedom, a winner of the National Medal of Technology, and he was inducted as a pioneer in the Internet Hall of Fame.
And if you still are not sufficiently impressed, let me tell you, he was named one of the 25 most intriguing people by People Magazine.
Thank you, Vint, for joining me today. It's great. It's great to have you.
VC: I have to tell you about that People magazine thing. That was 1994. I got one page. My wife, four years later, got six pages in People Magazine, and she doesn't have a PR department.
JR: Well.
VC: So, it's a small little thing in our household about that.
JR: Okay. Well, if we're starting informally, I'm going to start here. You're wearing a three-piece suit. Every time I see you, you're wearing a three-piece suit. And yet you work in a sector where hoodies are kind of standard. So, tell me about this sartorial thing.
VC: Actually, there is a story. My wife and I were at Stanford University, where I'd been working on the Internet, and with the support of the Advanced Research Projects Agency, now called DARPA for Defense Advanced Research Projects Agency. And in 1976, they said, okay, so you've been at Stanford working on the Internet. Why don't you come back here to Washington and run the program? And I said, no. And they said, you can't say no. So, my wife said, Washington, D.C., we've never been there. Three-piece suits. So, she goes off to Saks Fifth Avenue in Stanford and buys three three-piece suits, one of which is a seersucker outfit, because she knows that it's hot and muggy during the summer. So, I show up at work at the Defense Department in my three-piece suit. And not long thereafter, I'm asked to testify at some congressional committee. And I don't remember which one it is now. But I did my testimony wearing my seersucker three-piece suit. And I came back, and, you know, life goes on. And then a few weeks later, the director of ARPA asked to see me about my testimony. And I'm thinking, oh, God, I screwed up. It's the end of my government career. So, I show up, and he's in his office. He's got this letter. He says, I have a letter here from the Chairman. He says, thank you very much for Dr. Cerf’s testimony. By the way, he's the best-dressed guy we've ever seen from DARPA. And I took that as positive feedback, and I've been wearing three-piece suits ever since.
JR: Alright. Okay, now I'm going to ask you to roll back to the start.
VC: Alright.
JR: So, what is it about your upbringing that put you on the path to become, you know, a father of the Internet?
VC: I have to point to my fifth-grade math teacher as the person who really launched my interest. I was complaining to him about math being boring, you know, multiplication, division. There must be more to this than that. And he said, yes, there is. And he handed me a seventh-grade algebra book. I went home and worked all the problems in the book, and especially the word problems, which I love, because it was like an Agatha Christie thing. You had to figure out what's X. And so, I became a great math enthusiast. Then in high school, I got very excited about chemistry. And my best friend, Steve Crocker, and I got access to the computers at UCLA. So, I went to high school at Van Nuys High in Southern California, in the San Fernando Valley. So, all of this exposure just drove, you know, my career for all practical purposes.
JR: Yeah, I'm trying to wonder what the computers were like at UCLA then.
VC: One that we first got to program was called a Bendix G15, which some of you would say sounds like a washing machine. It was a numerical control computer that used paper tape for programming. And so, we would punch the programs in on the paper tape, feed it into the computer, wait for a while, and it would punch out a bunch of tape, which we'd take over to the Flexo writer in order to get it to print out the results. Then we got to use the 7094 at UCLA, which is a very large-scale, at the time, IBM machine. So, they were big, giant machines, batch processing systems. But then, in the early 1960s, Steve and I got to college. Timesharing was invented at MIT and came to Stanford around 1962. So, timesharing became the norm.
JR: Alright. So, tell us how you got from that paper tape to your breakthrough with TCP/IP, and how it really fit into what was happening with the early days of the internet.
VC: So, the background is that the Defense Department decided that it should be researching artificial intelligence way back in the 1960s. And, of course, you had . . .
JR: Way back in the 1960s.
VC: Yes. I mean, the program began in, like, 1962 at DARPA. And it was MIT and Stanford and Carnegie Mellon. And it persists to this day.
JR: Yeah, it's amazing because so much of the current conversation assumes it started with ChatGPT about a year ago.
VC: Yeah, right. It's actually 60 years of research and many phases and periods of dry spells and things like that. So anyway, ARPA is getting pretty excited about artificial intelligence and computing. And it starts to get interested in connecting universities together. The reason was economic. Every one of the computer science departments that they were funding to do AI research kept saying, 'You have to buy us a new world-class computer so we can keep doing world-class research.' And even ARPA couldn't afford to buy 12 computers, you know, that'd be like $6 million each, $72 million. Back in the 60s, that was a lot of money. So, they said, ‘We're going to build a network and you're going to share.’ And everybody hated that idea. And they said, ‘We're going to build a network anyway.’ Then the question was how to build it. And the technology of the day, of course, was circuit switching.
That's how the telephone system worked. But, you know, it's going to take forever for the computer to dial up a number, wait for the other computer to answer, send some data, hang up, then dial the next one. So, instead, this alternative. It was called packet switching. It's like electronic postcards and you just fling them into the network as fast as you can. And the network switches them. It's like the post office running a hundred million times faster. And so, the idea was to test that technology to see if it would work for computer communication. And by this time, of course, we're talking about timeshared interaction. So, people were doing real-time interactions remotely with the timeshared machines. So, we built the ARPANET based on packet switching and it worked.
Now, at this point, the Defense Department is saying, maybe we can use these computers for command and control. But if you're really serious about that, what does that imply? Well, command and control historically was voice communications and messaging. AutoDIN and AutoVon, for those of you who remember those dinosaur communication systems. So, we were looking at that point at voice, video, and data. Saying, how do we provide that using a computer communication system? So, we were experimenting with packet voice and packet video in the early 1970s. And of course, as the internet was the next obvious thing to do, because we had one network, the ARPANET. But we also needed a mobile capability. And that meant radio, because
you can't use wires, you know, they get all tangled up in the tapes, run over them. And you couldn't do ships at sea with wires. Because the ships get all tangled up and the aircraft can't get off the tournament. So, we had to build mobile radio systems and for long distance communication, satellite cones. So, we had a packet satellite network, a packet radio network, and the original ARPANET. So, Bob Kahn shows up in my office in 1973, literally just over 50 years ago. And he says, ‘We have a problem.’ And I look at him and I said, ‘What do you mean we?’ And he says, ‘Here's what we got. We got a mobile radio network. We got this packet satellite network. We got the original ARPANET. And we got to figure out how to hook them together. But we want to make it look all uniform. And oh, by the way, you can't change any of the networks to know that they're connected to anything else.
Each network thought it was the only network in the world. Well, the only solution to that problem is obvious. You need a box that knows how to talk to each of the two computers and can pass data back and forth. And you could say, does it have to translate stuff? And we looked at that. I said, that's a dumb idea, because once you get N networks, you're going to have, you know, N squared different translations. Never works. So, we said, everything has to look uniform. So, we invented the notion of an Internet packet and an Internet address. And, you know, the rest is history. So, the boxes in between knew about the Internet.
The host computers on the edges, the timeshare machines, knew about the Internet. The networks didn't know about the Internet. And so, when you see pictures, we would draw each network like a cloud. Because we didn't care what was going on inside. All we cared about was the boxes that connected the clouds and the computers that were hooked onto the clouds, which is where the term cloud computing comes from. We literally used to draw pictures that looked like clouds connected to each other. So, you know, that's where this all comes from. By the way, this is not just Bob and me. Eventually, a bunch of graduate students and then lots and lots and lots of other people helped to make the system what it is today, which is rule number one, if you want to do something big, get help, especially from people who are smarter than you are.
JR: Alright. So, as a federal official, I'd love to hear you talk a little bit about the role the government played as a catalyst for the Internet's development. And if any of those lessons still apply today. And I'm going to give you bonus points if you bring up Al Gore.
VC: Oh, Ok, ok. So, first of all, the U.S. government had everything to do with the Internet's evolution, creation and evolution, ARPA first, of course. But by the time we demonstrated the Internet capabilities in 1983, we turned it on, turned on the Internet. We had mobile radio, satellite and the ARPANET. And it wasn't long after that that the other agencies, NSF, DOE and NASA, each one of which had groups and communities to be interconnected. The labs at DOE, the labs at NASA, and 3,000 universities supported by NSF are all looking at how to interconnect, how to build a network to link things together. So, Al Gore, as senator, has a hearing in 1986 that Bob Cohn attended. Bob introduces the term national information infrastructure, which Senator Gore then translates into the information superhighway. And at the end of this hearing, he asked the question: should we build an optical fiber network to connect the supercomputers that DOE and NSF are sponsoring, five of them, to the rest of the academic community? He asked that question in 1986. So, we all went off to San Diego in February, because that's better weather down there than here, and came back with the National Research and Education Network Plan.
And that led to the NSF Net Backbone, which gets started in 1986, the same year that Senator Gore had that hearing. And so, NASA decides to build the NASA Science Internet. DOE decides to build the Energy Sciences Network. And NSF decides to build the NSF Net Backbone and a dozen intermediate-level networks to connect groups of universities together.
This was a brilliant move on NSF's part for two reasons. First, it took advantage of the fact that the Internet was designed to connect networks together. And second, by having a dozen of the intermediate networks, the NSF Net Backbone operators only had 12 customers to worry about instead of 3,000. And each of the intermediate networks could take care of their group of universities. So, NSF really deserves a lot of credit for testing the theory of the TCP/IP protocols. And to their credit, they chose TCP/IP in spite of the fact that there were giant protocol wars going on starting in 1978 when the Europeans decided they didn't want to have anything to do with this defense-sponsored TCP/IP stuff. They started a new initiative called Open Systems Interconnection that involved the International Standards Organization. And it was built on the bottom part of X-25 and X-75, which came out of ITU, C-C-I-T-T. Well, as far as we know, OSI died an ugly death, and it never quite made it out of the box. X-25 was around for quite a long time. The banking community picked it up. I used it when I built MCI Mail. I used the X-25 networks because I couldn't get IBM and Digital Equipment Corporation and HP to adopt and use the TCP/IP protocols because that was 1983.
We had just turned the Internet on. They didn't know what it was. So, I couldn't get them. They used that, but they knew about X-25. So, I used that. I actually ended up running an X-25 system when I returned to MCI in the 1990s, but I finally shut it down in 2003.
JR: So, a lot of this has been about the early days, the government and university culture that produced Internet networks. I'm just wondering, was there some eureka moment, maybe then or as we reached the commercial Internet, where you basically said, well, this is going to be big?
VC: It took me longer than some of my colleagues to figure this out. I'm kind of slow, slow out of the box. That's why I'm not a billionaire as opposed to some of my other friends. We turned the Internet on officially 10 years after we started the work. So, January 1, 1983, is the big transition. Everybody has to run TCP/IP on the ARPANET, packet radio net, and packet satellite net. And in 1984, Cisco Systems gets started to build commercial routers. And who will they sell to? They'll sell to the academic community, which wants to build pieces of Internet on the campus so they can then hook the campus up to the backbone networks. And, oh, incidentally, in 1973, the year that Bob and I start working on the Internet design, Bob Metcalf and David Boggs are at Xerox Palo Alto Research Center, and they invent the Ethernet, which is a local network. And Sun Microsystems comes along and builds workstations that run TCP/IP because that was the common academic standard at the time. All this stuff is happening in the early 1980s. So, what you're seeing is an acceleration of commercialization until 1988. Internet is only permitted to be used by academics and people with Defense Department contracts. And I walk into an exhibition that was called Interop, that was Interoperability. Dan Lynch was the founder of this exhibition. And he basically said, you have to show that your software and hardware works with everybody else's, or I'm not letting you into the show. So, he’s got this giant yellow Ethernet cable snaking through the exhibition hall. And everybody plugs into that.
And all kinds of debugging take place, you know, the day before the show opens. So, it starts in 1986 with just lectures. Then in 1987, it's an exhibition. In 1988, 50,000 people show up at Moscone Center. And I walk in with Eric Benimo, who is then the CEO of 3Com, the company that commercialized Ethernet, that had been started by Bob Metcalf. So, I see this giant two-story Cisco dashboard. It's big enough that there are people on multiple stories and equipment on display. And I turned to Eric, and I said, ‘Eric, how much do those cost?’ And he says, ‘About a quarter of a million dollars.’ Remember, this is 1988. That was a lot of money. That doesn't count the cost of the people to man the booth for a week. And I just stood there, you know, my jaw is on the floor, thinking, my God, people must think they're going to make money out of the Internet. And then I realized, well, how the heck is the general public going to get access to this? There's an appropriate use policy at NSF that says only government traffic, only research traffic on the backbones.
And I'm thinking, well, that's hard to, you know, you can't get the private sector. So, what am I going to do? So, I had built MCI Mail as a commercial email service. So, we went to the Federal Networking Council, which had policy responsibility for the Internet. And we said, is it okay to connect the MCI Mail system to the Internet as an experiment? Now, to see if we could make it inter-work. Of course, my real purpose was to break the appropriate use policy by allowing commercial traffic to flow on the backbone. And they said yes. So, by '89, we get it running. We announce this. Every other email carrier says we have to get in on this. CompuServe wants to get in on time. Telemail and all the others insist on getting onto this Internet thing. What do they discover? They discover first, they can make it work. And second, all of their customers can talk to all their competitors' customers because now they're compatible through the Internet. Holy moly, what happened to our little island? 1989, by the fall,
free commercial Internet services pop up. UUNet, PSINet in Virginia, and CerfNet out in San Diego, which was originally supposed to be spelled S-U-R-F-Net because what else would you do in San Diego but surf the Internet?
They had T-shirts and the whole thing all set up. And then somebody discovered that somebody in the Netherlands had already named their network S-U-R-F-Net, connecting research networks in the Netherlands. So, they decided to rename their organization the California Educational Research Foundation Network, C-E-R-F-Net. Then they called me, and they said, 'Is it okay if we call it CerfNet?' And my first reaction was, 'If they screw it up, am I going to be embarrassed?' And they said, 'Wait a minute, people name their kids after other people, and they don't blame the people they named them after if they don't come out right.' And I said, 'Sure.' So, I flew out with Susan Estrada, who was the executive director. We took a plastic bottle of glitter, and we smashed it over a Cisco router, and we launched CerfNet in 1989.
JR: It worked out okay.
VC: I know these are all . . . well, it was eventually acquired by AT&T. And by that time, I at MCI as senior VP for engineering. And I call them up and say, can I have my name back? And they said, no. So much for that. I'm sorry this goes on and on. You know, you turn the spigot.
JR: So, if you go back to those days, is there something that, looking at the world as we know it now with the Internet, that you wish you knew back then?
VC: Oh, yeah. When Bob and I were doing this original design, we were trying to figure out, okay, how much address space do we need? I mean, this is 1973. Computers are hoarding big things that are expensive. So we said, well, maybe there'll be two networks per country. So, there'll be some competition. We didn't know how many countries there were. We didn't have any Google to ask. So, we guessed at 128, because that's a power of two. And we doubled that. That's 256. Eight bits of network address. Then we said, how many computers? And we thought, you know, how about 16 million? Remember, these were big, expensive time-sharing machines. But we said, what the heck? 32-bit address space. That would be enough, if densely packed, to handle 4.3 billion computers on 256 networks. And that was more than there were people in the world at the time. Now, I would have to whisper in my 27- or 30-year-old ear, saying, you need 128 bits.
And I'm sure I could not have sold that to anybody. That's 3.4 times 10 to the 38th addresses. And I used to think that was enough address space to let every electron in the universe have its own web page. Until somebody at Caltech sent me a note, ‘Dear Dr. Cerf, you jerk. There are 10 to the 88th electrons in the universe, and you're off by 50 orders of magnitude.’ So, I don't say that anymore. So, I would have said that one thing, you know, bigger address space, but I couldn't have sold it to anybody. And the
other one might have been more security. However, remember, at that time, the Internet is being built by a bunch of graduate students. They are not the cohort I would turn to for discipline in terms of key management. So, trying to do that probably would have prevented the Internet from even happening at all.
JR: So, one thing I know you've talked about with respect to the Internet a lot is just the need for open architecture and how that was the key to its evolution.
VC: It's true.
JR: So, I'd love it if you talk about, you know, policies to protect Internet openness and why that's so fundamental to the networks you help with.
VC: Well, if you look at the Internet architecture, it's a layered structure. You know, the lowest stuff is physical communications and twiddling bits. And then there's framing. And then there's the Internet protocol. And then there's TCP and QuasiCPUT. And then there's FTP applications and so on. So, this layering was infinitely extensible. You could add new protocols at any layer. So, when optical fiber came along, we just swept it in and carried IP packets on top of optical fiber. When Tim Berners-Lee came along in 1991 and said, 'I want to add a new layer of protocol called Hypertext Transport Protocol,' we said, 'Great, no problem. Just stick it right there on top of TCP/IP.' And he did. And he got immediately a global platform to run the World Wide Web. So, that was a smart move on his part. And it was great for us because it made the Internet more useful. So, the openness of the protocol architecture was very important because new contributions could be made freely. And second, the connectivity is everything in this system. When we started out, we said everything on the Internet should be able to talk to everything else. You don't have to answer, just like phone calls. You don't have to answer the phone, but you want to be able to call anybody. You can hang up if you don't like the conversation. Same is true on the Internet. But we wanted everything to be connectable because we didn't know at the time we were doing the design, what would need to talk to what. We knew it was going to be computers talking to computers. And by the way, we were very egalitarian. We said, ‘You know what? A supercomputer should. It should be able to talk to a workstation. And it should be okay. It's just that the supercomputer may have to not talk as fast as it's capable of talking. But that's what flow control is for. And that's what TCP does.
JR: It's a very democratic network.
VC: That was with a lowercase d, yes.
JR: Lowercase d, yeah. Alright. Let's talk about AI. You brought it up early. But in the current environment, you know, we've got people talking about how it can improve our lives. We've got people talking about how it can threaten our very
survival. So, in this environment, tell me, are you an optimist or a pessimist?
VC: I'm a techno-optimist. And so, AI is not going to do us all in. I really don't think that. It's not an existential threat. If you're looking for an existential threat, try global warming. That's an existential threat. AI is a really interesting challenge. How to take this technology and understand it well enough to fashion it into the kind of tool that it can be.
JR: So, what do you think government should be doing to seize those opportunities and avoid pitfalls?
VC: So, the one thing the government should not try to do is to regulate the technology itself. I think that would be a mistake because the legislators and, frankly, even the experts don't fully understand this stuff. They certainly don't understand it well enough to regulate the basic technology. But what you can do is say, well, how is this stuff being used? And you can start looking at applications and you can ask, okay, is this a high-risk or a low-risk application? If it's a high-risk application, like medical diagnosis, treatment, financial planning, psychological counseling, you should ask the question as a legislator, that sounds high-risk. Please show me how you have mitigated the risk to the users of those applications. Demonstrate your due diligence. Now, if it's a kind of a low-risk thing, for example, entertainment, then you don't have to demand quite as much. But I would say that looking at applications, we should legislate the risk mitigation primarily. And over time, as we understand these things better, the risks will probably come down in terms of their use. Certainly, in terms of medical diagnosis and things like that, we're still seeing very effective, if narrow, applications that actually do seem to work better than doctors do.
JR: Okay. So now you've established yourself as a digital optimist. It's a posture I share. But I also know, when I was doing some research before our conversation, that you've raised concerns about what you call the Digital Dark Age.
VC: Yes.
JR: So, you're going to ask me to talk a little bit about something.
VC: So, actually, you should go back to AI for just one second. All of you are familiar with the most recent manifestation called large language models. This is generative AI as opposed to analytical AI. Generative AI does have some problems because we don't fully understand how it generates its output, and we can't predict when it's going to generate counterfactual output, even if it was trained on factual material.
So, to test this, I decided to ask one of the bots to write an obituary for me. And, well, I figured that obituaries are on the net. It would have ingested the format, you know, so it was reasonable to assume these bots knew what an obituary was. And there's
information about me on the net. So, I figured that was a reasonable question. And it produced a reasonable, you know, 700-word obituary. We're sorry to tell you Dr. Cert passed away. And then they put a date in, which I didn't like very much. And then it got to my career, and it got to family members, and it conflated what I did with what other people did. And it made up family members as far as I know. I said, well, how could that happen? And the answer is lack of context. So, these things build these statistical models of how words appear in sentences across a huge amount of context, and then it generates output. But it doesn't know that your bio and my bio happen to be on the same web page. But it doesn't know that this. This paragraph is you, and this paragraph is me. So, it could be plucking words from various places, putting them together, looking grammatically correct, even persuasive, and dead wrong. So that's our big problem with generative AI right now, is understanding why it produces these odd results. You know, I look a little bit like Sigmund Freud, and I'm thinking of doing a video. Well, what we have now is the artificial it and the artificial ego. But what we are missing yet is the artificial superego. The control all the uncontrollable impulses of the artificial it. Well, sometimes a cigar is just a cigar. So, we have work to do. But please let us do the work. That's what I think the legislators and the regulators need to let us do the work to understand how to fashion these tools more effectively. So there. You had another question that you wanted to ask.
JR: What do you mean when you from that when you referred to the Digital Dark Age? Is it the element of AI?
VC: No, no. This is an even more fundamental problem.
JR: Okay.
VC: I'm sure you all will remember this. You know, 6,000 years ago, when people were doing inventories and things, they used clay tablets and a form of writing called cuneiform, Sumerian. And those clay tablets in the warehouses often were preserved for thousands of years because the warehouses burned down and baked the tablets. Now, if you can still read cuneiform, you can still read this stuff. And then we invent vellum, which, you know, is calfskin and things like that lasts for 1,000 to 2,000 years. You notice that's less than 6,000. That lasts for maybe 1,000 years, 500 to 1,000 years. You notice the numbers are getting smaller. Then comes book paper. Then comes newspaper. Okay. That stuff lasts three days. Now comes digital technology. Oh, boy. Fantastic. Look at how many bits we can put on this 5 1/4-inch floppy disk, 800,000 bytes. And then 3 1/2-inch floppy disks come along. How many years did the 5 1/4-inch floppy disk last? And can you still read the bits on it? The answer is the bits may still be there, but you can't find a reader to read it. I found one on eBay for a 3 1/2-inch floppy disk because I had a pile of them. I put them in the closet, and I wonder what's on there. I found a bunch of files, and I could read them on my Mac, but they were WordPerfect files, and I didn't have the software
to run WordPerfect anymore. So, the problem of the Digital Dark Ages is the loss of digital content because either we can't read the medium anymore or the medium disintegrated like polycarbonate DVDs might; last 10 years or 15 years or the software that was needed to understand the format of the bits isn't running anymore on the operating systems of the day. So, the Digital Dark Ages is where we lose all kinds of access to digital content, including imagery and programs like games or spreadsheets and text materials. And that's my big worry, that we have to institute some kind of preservation regime that hangs onto formats, operating systems, application programs, and the ability to correct and interpret the bits on digital media.
JR: That's so interesting. Alright, let's talk about something else, though. In 2023, this agency created a new Space Bureau because we took notice of all the communications work in our sky, just multiplying so fast. And I know that you've often talked about the interplanetary Internet.
VC: Right.
JR: So, I'd love it if you could explain what your vision is for interstellar communication. Well, we can't.
VC: Well, let me just start with interplanetary is pretty hard. But we started working at the Jet Propulsion Laboratory at NASA in 1998, 25 years ago, just after the Pathfinder mission landed successfully on Mars. Some of you will remember in 1976, the Viking 1 and 2 landed on Mars successfully. Then nothing worked for 20 years. It missed the planet, crashed and burned. And then in 1997, Pathfinder lands with a little rover. And we get data back. So, I got so excited about that. I got together with the Jet Propulsion Lab guys and said, ‘What should we be doing now?’ That is to say in 1998, that we're going to need 25 years later? And we all looked at each other and said, an interplanetary Internet backbone. So, we started working on the design, and we immediately discovered the TCP/IP doesn't work with a 40-minute round-trip time. It takes 20 minutes at the speed of light, to get between Earth and Mars when we're farthest apart in our orbits, and 20 minutes to go back. It's only seven minutes round-trip time when we're closest together in our orbits. But it gets worse when you go to the outer planets. There's another problem, too. Planets are rotating, and we don't know how to stop that. And so, if you're talking to something on the surface and the planet rotates, you can't talk to it until it comes back around again. So, you have variably delayed and disrupted communication. We started designing new suites of protocols for delay and disruption tolerance, and networking, and that's what we have now today. It's called the Bundle Protocol Suite, and it deals with those problems that TCP/IP couldn't deal with. So, we're in the process of deploying. We've been on the International Space Station for over a decade. We have prototype software that's been on Mars since 2004, supporting all of the subsequent American missions that have landed on Mars since 2004 have been running the prototypes. So, we're part of the Lunanet plans. We're part of the Artemis mission, the return to the Moon. So that's interplanetary networking, and we're confident that we can do that. Then we did start talking about interstellar. We said, okay, got the solar system solved. Now what about Alpha Centauri, 4.3 light-years away? And so, a couple of problems there. First one is that current propulsion systems would take 76,000 years to get there. It's kind of a long time for an experiment.
JR: I'm not going to fight for that myself.
VC: Yeah, well, I think—that's seven times longer than our civilization has existed. So, you know, that's a problem. So, we need better propulsion to get up to 10, 15 percent the speed of light. We'd like to get there in 100 years, basically. Then there's the navigation problem. You know, when we send missions out to the solar system, in the middle of everything, you do mid-course corrections. You say, where are you? Oh, you know, burn for 30 seconds in this direction. Well, if you've got a spacecraft that's a light-year away, it takes a year to send it. It's a signal saying, do a mid-course correction. It takes another year to find out what happened. This is not exactly interactive. Fortunately, you know, we have this global positioning system.
It turns out that there are stars in our galaxy and other galaxies that are pulsars whose location is not going to shift very much, even if we move 4.3 light-years away from Earth. And so, they can become a galactic positioning system as opposed to a geocentric positioning system. And so, we think that navigation could be solved on board, in effect, by using the pulsars. Then there's one other problem. How do you communicate with something that's 4.3 light-years away? And we think, well, lasers would be good because that's higher speed than radio. And, of course, we have to have enough power so that the signal is actually detectable 4.3 light-years away. But here's the big problem. You have to think for just a second what it means to be 4.3 light-years away.
It means that the signal that you're going to send is going to take 4.3 years at the speed of light to get here. The problem is, where is 'here'? If you're at Alpha Centauri, which direction do you point so that your signal gets to where Earth is going to be 4.3 light-years later? That's a non-trivial exercise. So, not being smart enough to do this ourselves, the interplanetary system, the Interstellar Network Special Interest Group of the Internet Society went to Lincoln Laboratory and said, could you do an analysis to tell us how far away we are from the engineering point of view to being able to mount a mission to Alpha Centauri? And they came back and said, you're off by a factor of 53 dB. And if you don't speak dB, that's a factor of 100,000. So, we still have a lot of work to do to get to Interstellar.
JR: Alright. So, I'm going to go from the heavens to something much more local here in Washington. I know you've served on the board of trustees of Gallaudet University, which is our neighbor here. I've spent quite a bit of time at Gallaudet and with the
president myself. And of course, this is the university that is dedicated to the education of deaf and hard of hearing students and is an extraordinary gem. So, I'd love it if you could talk a little bit about your service there and what we can do to make the Internet more accessible for people who have a broad range of disabilities, including those.
VC: Thank you for asking that. I'm a big fan of accessibility and using the technology of computing to improve the accessibility of these tools to people with various disabilities. First important observation is all disabilities are different. They may fall into a category like visual problems or hearing problems, like I have, or motor problems, but every person is individual and different. So, the accommodation that's required has to be flexible. The people who write user interfaces don't have a lot of lived experience with the disability. And so, their intuitions are not necessarily all that good. They need examples of what worked and what doesn't work so they can begin to intuit that. So, we have classes at Google where we teach people about these things. But it does turn out that there's still a lot of work to be done to help people, programmers in particular, user interface designers to understand how to do that. So, there's still a lot of work to be done.
If you don't mind though, I have a story to tell about technology and its role in helping people with disabilities. It's not about me and my hearing aids. I've been wearing them since I was 13. It's my wife and her two cochlear implants. That's the big deal. She was born with normal hearing, but at age three, she lost her hearing to spinal meningitis. All the little cilia hairs in her cochlea just burned up. So, she was totally deaf and learned to lip read at the John Tracy Clinic in Los Angeles, and then spent her entire school career lip reading in high school and college and in elementary school.
We met in Los Angeles at our hearing aid dealer because she was wearing a big Sonatone body aid to just get a little bit of sound. She was getting some low frequencies, that was it, but mostly she was lip reading. So, we moved here in 1976 so I could go to work at ARPA, and she was such a good lip reader. I thought she was working at CIA but couldn't tell me about it. Then, in 1996, 20 years after we moved here, she learns about cochlear implants, gets on the internet, and talks to Johns Hopkins University where they were doing cochlear implant surgery. And so, John Naparko, who's sadly passed away now, but did both of her cochlear implants, the first one in '96, the second one in the other year in 2006.
This is really amazing. It's an outpatient operation. 45 minutes later, you've got a cochlear implant. You go home for a couple of weeks while everything heals. Then you go back to be activated, which sounds kind of religious, but what it really means is that, that you turn on the speech processor, put in a map. This thing takes sound in, analyzes which frequencies are present with a Fourier transform, figures out what their amplitudes are, and then maps that to the electrodes that are embedded in the auditory nerve. The brain is getting neural stimuli from those electrodes, which it interprets as sound. Now because she had not lost her hearing until she was three years old, her brain had already discovered sound and knew what those signals meant.
30 minutes after her first activation, she picked up the phone and called me. Now we'd been married 30 years at that point and had never had a phone conversation. So, it wasn't a very deep conversation, but it was important.
JR: But no phone here. We like that at the FCC.
VC: Well, so I get home, and I got a 53-year-old teenager at home. I can't get her off the phone, but she'd take any call that was coming in. One of them was AT&T. Now I'm senior VP of engineering at this point at MCI. AT&T calls and she says, ‘HI, you know, where are you? Oh, India. Well, your English is really good. Where did you learn that?’ She goes on for about half an hour. This poor soul on the other end says, You're going to switch it out, aren't you?’ And she says, ‘No, my husband works for MCI, but thanks for calling.’ You know, she hangs up. Then, she decides that she wants to hear words that she hasn't heard before. So, she calls the library. Remember, she's on the phone calling the library. I'd like to sign up for recorded books for the blind. And they say, fine, no problem. Name, address, phone number. Now you're blind, aren't you? She says, ‘No, I'm deaf.’ And there is this long pause while they're trying to think about how's that going to work. She's determined that no decibel will go undetected. She gets little patch cords with microphones on them to clip to people's lapels. She gets FM transceivers so she could come to a place like this with our little microphone. She'd be sitting anywhere 150 feet away and she'd be getting perfect sound. She’s got patch cords to plug into the airplane television shows so that the only sound she's hearing is coming from that. No acoustics. And the three-year-old is screaming next door. She doesn’t hear at all. So, she’s very aggressive about using these technologies. And it’s such a fantastic example of what technology can do for accessibility. And we haven’t done enough on the computing side to make things more accessible. We need to work harder.
JR: That’s a wonderful story. There’s so much that I’ve seen here that technology can do for people who have disabilities. But I also think that we can learn so much from people with disabilities about how to use technology.
VC: Absolutely. I mean, the computer is the most flexible instrument ever built. I mean, think about it. It's an infinite number of things you can do if you can figure out how to program them. So, we haven't come close to doing all the things that we could and should do to make these things much more accessible or to help them be transducers for accessibility in other contexts.
JR: So last question before my final questions. I've seen you describe yourself as a kink orthodox. I love that phrase. So, what does it mean to be a member of this faith?
VC: Right. Well, first of all, that term didn't occur to me until I was lecturing in Moscow in 2010. And I did four evening lectures in public. And at the end of the lecture, at each one of the four, some different person got up and said, ‘Dr. Cerf. Do you believe in God?’ Because they knew that my title at Google is Chief Internet Evangelist. That wasn't the title I asked for. Larry and Eric and Sergey, when they hired me, they said, ‘What title do you want?’ And I said, ‘Archduke’. And you know, they went away, and they came back, and they said, you know, the previous Archduke was Ferdinand, and he was assassinated in 1914 and it started World War I. Maybe that's a bad title. Why don't you be our Chief Internet Evangelist? They said, okay, I can do that. So, I get asked this question and I'm sitting here thinking, how do I answer this question? And I said, ‘I'm geek orthodox.’ And you know that sort of fitted into the Eastern religion. They all understood and got the point. So, geek orthodoxy basically says, I really believe that technology is our friend here. Even when it doesn't seem to be, it can and should be. And that's my job is to try to make those technologies much more useful for everybody.
JR: Oh, a good note to end on. But I first want to ask you some of the questions we ask everybody in our first conversations,
VC: Uh-oh!
JR: Here you go. They're like rapid fire. Do you remember the first email you sent?
VC: No, I don't. I just have no ...
JR: That's incredible. I thought that you, of all people, would absolutely . . .
VC: Listen, listen. Everybody asked Ray Tomlinson, who invented networked email, what was the first one you said? And he said, ‘It was probably blah, blah, blah, blah, blah, blah, blah, blah,’ because he just typed some crap from the keyboard and sent it. It didn't matter what it said. Unfortunately, unlike Samuel Morse, 'What hath God wrought?' All I could think of was what wrought hath we got, but we didn't even do that.
JR: Alright. What's the first thing you do in the morning?
VC: I brush my teeth, and then I take three pills, and then I check my email.
JR: Alright. What was your first concert?
VC: Oh, gosh. I was trying to remember the composer and the piece. I can't. I'm sorry. But probably it was Brahms, although I'm a huge fan of all three Bs. I prefer classical music, Brahms, Beethoven, Bach, and so on. But I don't remember which one it was. I'm sorry.
JR: Definitely classical, though.
VC: Oh, yes. I mean, nothing after 1850, you would think.
JR: One of the fathers of the internet does not listen to music after 1850.
VC: Well, I mean, why do you think that? Why do you think I wear a three-piece hat?
JR: Unpack that.
VC: I love watching Sherlock Holmes movies because they're all wearing three-piece suits.
JR: Your people. Okay. What's the one bit of advice you'd give to someone on the first day of their first job?
VC: Be humble and ask a lot of questions.
JR: That is wonderful advice. Alright. Finally, this podcast and speaker series celebrates those who paved the way for others. So, I'm going to close by asking, can you tell us about a mentor or a mentor who influenced your work?
VC: Well, there are several people to whom I owe my career. My math teacher in the fifth grade, for sure, Mr. Tomaszewski; Bob Kahn, who has been my mentor ever since we met in 1970; Steve Crocker, my absolute best friend from high school, whom we met in 1959. Those people have really made a huge difference, and my thesis advisor, Jerry Estrin. Those were all people have made a huge difference for me in my life, and many of them are still around and still helping me be a better person.
JR: Oh, I think it's always wonderful to call them out, so thank you for doing that. Alright. I feel like this is appropriate. Where can folks follow you on the internet to keep up with what you're doing?
VC: So, I write a column every other month in the communications of the Association for Computing Machinery, but most of you probably are not subscribers to that. I have a website which is not yet functioning. It's going to be surfsupp.com. And so, the column is Surf's Up at ACM. I have somebody helping me organize that, so stay tuned, surfsupp.net will be coming soon.
JR: Oh, something to look forward to. Well, that wraps up this edition of First Conversations. Thank you for being here, Vint, for telling stories, telling us about the work you do, and thanks to everyone for listening. Take care.
VC: Well, thank you for all the great questions. And thank you, folks, for sitting patiently and listening.