Vint Cerf codesigned the foundational architecture of the internet. He’s a renowned computer scientist, Google’s chief internet evangelist, and today’s guest on this episode of the At the Edge podcast. Cerf speaks with McKinsey senior partner Lareina Yee about the genesis of the internet, the need to improve online accessibility, and the importance of accountability.
This transcript has been edited for clarity and length. For more conversations on cutting-edge technology, follow the series on your preferred podcast platform.
Sketching the internet on an envelope
Lareina Yee: Let’s go back to 1973. You were working on connecting ground-based military computers to communication satellites and mobile radios. You had an envelope on which I believe you sketched the architecture of the internet. Could you tell us a little bit about that time? What was your original vision?
Vint Cerf: The original vision came from my colleague Robert Kahn. We’d gotten to know each other because he was involved in a predecessor system called the ARPANET [Advanced Research Projects Agency Network], which was designed and built for the Advanced Research Projects Agency [ARPA, now DARPA] to connect research universities together that were being funded by ARPA for artificial intelligence and computer science research.
He had the idea that computers could be used in support of command and control. But that implied dealing with mobile vehicles, such as ships at sea and aircraft. You can’t do that with dedicated wires, so he had already begun working on a mobile packet radio and packet satellite system to combine assets and support that kind of wireless communication. We also realized we needed to support voice, video, and data as part of the command-and-control challenge.
We then spent about six months working together in 1973 to try to figure out how you could build a system that would allow multiple networks to interconnect with each other in a very uniform and smooth way, even though the insides of each network were quite different. The packet satellite system had different error rates, data rates, and certainly different latency rates. The packet radio system had dynamic connectivity, and you had to deal with that.
The result of our six months of work was the design of the internet. The little sketch you referenced was my effort to visualize the results of our discussions. What came out of all that was actually a fairly simple concept of linking networks to each other with things we called gateways, which today are called routers. The host computers, where all of the interesting stuff happened, were outside of each network but needed to know that they were part of this multinetwork system. The end result was the internet protocol suite.
Lareina Yee: Can you tell me about that germ of an idea from a more personal level?
Vint Cerf: Part of the motivation for developing the internet was the success of the ARPANET, and a number of applications were developed for that system. One of them was remote access to time-sharing systems, so students doing research at one university could access the computing resources and results of students at others. And that was an effective way of accelerating the pace of research.
Second, electronic mail was developed in 1971 on the ARPANET, which became a very popular application. But it’s important to realize that although we were working on the technical aspects of this interconnection, many, many other people were involved in thinking through possible applications as well as implementation and operations. This was a huge collaboration.
The problem of online abuse
Lareina Yee: When you reflect forward and think about other new technologies, what are some of the important lessons—not just from a technology perspective but also from an ethical standpoint—as something that began as a sketch becomes such a profound reality?
Vint Cerf: The thing I would note most is that once you become dependent on something like the internet, you would like it to be very reliable. And so, reliability is always in the back of my mind as a very critical and desirable outcome.
Second, you want resilience, so if something breaks, you don’t lose access. And third, you want it to be affordable. That means the business model has to be sustainable. All of those things are very important, because once you become dependent on them, it’s not clear what happens if you can’t count on one of them anymore.
You also mentioned ethics. And I have to say, one of the disappointing aspects of the internet’s evolution has been online abuse, whether that’s bullying on social media, ransomware, malware, or the rapid spread of misinformation and disinformation.
One of the disappointing aspects of the internet’s evolution has been online abuse, whether that’s bullying on social media, ransomware, malware, or the rapid spread of misinformation and disinformation.
This system is an amplifier of access to information. So figuring out how to winnow the wheat from the chaff is still a very important problem. And there’s not a purely technological solution to this problem. There is also this thing called wetware, which is our brains.
And I think people need to be trained in thinking critically about the content they are being exposed to on a daily basis. I’ve even given some thought, just half jokingly, to an internet driver’s license, where people have to learn how to be safe in the online environment.
The need for accountability
Lareina Yee: Let’s think about that driver’s license. What are the components of that test?
Vint Cerf: There are a few things that I’d like people to know just so they have some idea of how this thing works. The more important thing for them to learn is how to behave ethically and be safe online, which means not putting personal information in places where it doesn’t belong, not sharing passwords or other kinds of authenticators, and being somewhat skeptical of online claims.
We have work to do in the community to provide better provenance for where information comes from so that people can use thoughtful judgment about whether to trust the information or not.
Lareina Yee: As you think about that internet driving license, given the volume of use and maturity of the technologies, how do you put the genie back in the bottle?
Vint Cerf: I don’t think it’s possible to put the genie back in the bottle. So the real problem is, how do we manage the genie? I think there are probably social norms that have to evolve in our community.
Anonymity and anonymous speech are often valued in democratic societies. But I’ve become persuaded that anonymity must include accountability, and that with the benefits we get from this online environment come responsibilities—and we need to recognize that.
Particularly, harmful behavior needs to be addressed, and it needs to be addressed on an international basis. Because the victim may be in one jurisdiction and the perpetrator in another, which means we may need international agreements in order to cope with that kind of problem. A cybercrime treaty has been negotiated that is trying to address exactly those kinds of problems.
As a society, I think we need to emphasize and value accountability for online behavior, while at the same time doing what we can to protect people’s privacy and safety.
Parallels between internet governance and climate change
Lareina Yee: When you think about how international cooperation or governance has supported technology in a positive way, are there analogies?
Vint Cerf: As all analogies have their weaknesses, I think we have attempted to do things that I would say are similar. For example, in response to global warming, there are attempts to cooperate, recognizing that we have a shared environment called the atmosphere and the oceans. The internet is also a shared environment, so we might draw some parallels.
In response to global warming, there are attempts to cooperate, recognizing that we have a shared environment called the atmosphere and the oceans. The internet is also a shared environment, so we might draw some parallels.
In shared environments where our actions affect others, we need to adopt practices that recognize this and improve the environment for everyone, whether that’s avoiding more carbon dioxide generation, avoiding ocean pollution, or trying to reverse the growing plastic and microplastic pollution. All of those sorts of things require collaborative work among countries and peoples, and I think the internet is no different.
Improving online accessibility for all
Lareina Yee: You’ve been a leader in thinking a lot about the societal aspects of the internet. We’ve touched on some of these very big topics. Another one I think is deeply personal for you is accessibility. What do you think needs to be done to ensure that the internet and its relative services are truly inclusive and accessible?
Vint Cerf: It’s going to take a lot of work to make the internet, its applications, and all the other computer-based systems more accessible. There are a couple of reasons why that’s so hard.
The first one is that the people who design the software and user interfaces may not have a lot of intuition into what makes something accessible for a person with a vision problem, a hearing problem, a motor problem, or a cognitive problem.
The other problem is that whatever infirmity you might have, it can encompass a very broad spectrum. So people with hearing problems are either totally deaf or may just need a little amplification. And people with vision problems, similarly, could require magnification or a spoken rendering of what they’re theoretically looking at, because they can’t see it.
All of these things have many gradations, which makes it very hard to design software that works well across these broad spectra. So two things occur to me. The first one is more training and exposure to solutions to these problems, so you begin to see how things get solved effectively.
We also need people with those disabilities helping us try things out, test things, and make recommendations, because no one person with a particular disability can speak to the entire spectrum.
But there is some hope here. With the artificial intelligence and machine learning mechanisms becoming more and more feasible, it’s possible that these mechanisms will help us adapt to improve accessibility.
A trivial example of this is how powerful speech recognition has become so that we’re getting transcripts or captions automatically. The next obvious thing, of course, is language translation. I’m language impaired in many respects, as I only speak one or two, and there are literally thousands of them.
Those kinds of things can help, and that’s just on the language side. I think we’re likely to see other adaptable mechanisms arising out of machine learning that will permit us to build user interfaces that can be tailored to the needs of a particular user.
Lareina Yee: If you look back at the rapid proliferation of the internet and the quick spread of generative AI applications, what are some lessons that you would emphasize for us going forward?
Vint Cerf: First of all, because we’re making it easier and easier to build new applications, there will be more of them. There must be more than a million or maybe even two million applications for mobiles these days. What that means, though, is that just because there is such a proliferation of applications, there is not the same proliferation of skill, with regard to accessibility.
However, if we get really good at using machine learning to make user interfaces more adaptable, then that might actually help solve the problem. Because the users themselves would be engaged and interacting with the software, trying to get it to adapt and adjust—for example, color contrast, sound levels, or even selective frequency amplification. Things of those sorts could be done using machine learning tools.
Lowering the cost of online access
Lareina Yee: We’ve touched on some pretty big societal challenges, like accessibility, security, and cost. Technology innovation is often the solution to some of those challenges. As you look forward, what are some of the things that still need to be discovered or worked on?
Vint Cerf: One of the problems with access—not accessibility in the disability sense, but just access to the internet—is the cost can be fairly high for some types of access. Stringing optical fiber networks is not an inexpensive proposition.
One of the interesting developments of the recent past, of course, is the Starlink service coming from SpaceX, which has the ability to reach virtually any place on the planet. SpaceX hasn’t yet launched all the satellites that are planned, and the number is something like 30,000. If they all get up there in that constellation, it would be pretty hard to avoid access to the internet.
On the other hand, whether it’s affordable is a different story. That will vary depending on where you are. So finding ways to drive cost out for access to the internet is pretty important.
I hope we’ll also see community efforts. For example, there have been disputes over whether a community can build its own internet service. Some people say, “That’s the government competing with the private sector.” My reaction is, maybe it isn’t. Maybe it’s the community doing something that the private sector didn’t want to do.
So having these various ways of driving cost out and making things more useful and accessible seems to me part of our future.
The challenge of building an interplanetary internet
Lareina Yee: One of the many things you’ve worked on is communication networks in space. Can you talk a little bit about the interplanetary internet?
Vint Cerf: The team I’ve been working with on this began this endeavor in 1998 at NASA’s Jet Propulsion Laboratory, so it’s 26 years and counting. One thing we discovered very early on is that the terrestrial internet’s design does not necessarily work very well when you’re talking about interplanetary space.
The reason for that is because the distances between the planets are literally astronomical, and the speed of light is simply too slow. Just getting from Earth to Mars by radio—which moves at the speed of light—takes anywhere from three and a half to 20 minutes, one way.
Subscribe to the At the Edge podcast
There’s another problem too. The planets are rotating, and there’s no way to stop that. So if you’re communicating as a planet rotates, you have to wait until it comes back around to continue the conversation. Or maybe wait until a relay satellite comes around to pick up the signal and relay it.
We’ve also found that if you’re trying to communicate between, say, Earth and Jupiter, you have to go through the Mars relay. And it could be that when the data gets to Mars, it doesn’t yet have a connection to Jupiter and has to hold onto the data until the Jupiter link comes up.
That’s not how the terrestrial internet works. If there’s no path for the data to reach the destination, the router throws it away, believing you can retransmit it from the source. But in the space-based environment, it’s less likely that you’d want to do a retransmission all the way from Earth.
So the idea is to store information in the network, which is quite a different approach. We call this delay/disruption-tolerant networking. The rules governing that are called the bundle protocol suite and have been standardized by the Consultative Committee for Space Data Systems and by the Internet Engineering Task Force.
The various teams at the European Space Agency, the Japanese space agency, the South Korean space agency, and NASA have all been working together to implement and test these protocols. We use them on the International Space Station, and the South Koreans use these protocols on an experimental orbiter circling the moon. We’re anticipating being part of the Artemis return-to-the-moon mission as well.
The idea here is to design and build a space-based internet network that can be implemented by both governments and the private sector in the same way the internet was originally funded by government but now hosts major commercial activity as well.
We’d like to see the same kind of capability evolve for the interplanetary backbone network so it can support research for manned and robotic space exploration but also the commercialization of space.
This introduces all kinds of interesting questions into the policy arena. For example, are you allowed to own property on the moon? At the moment, the answer is no. But I think the commercialization of the moon may lead to questions like, “Why can’t I own the mine that I’m operating on the moon?” All those questions are going to need answers.
Developing an internet for nonhumans
Lareina Yee: At the risk of asking you even more, let’s say that you could have a team of scientists and engineers work on one or two other things. What would they be?
Vint Cerf: I do have a group that’s working on another interesting project. It’s called the interspecies internet. Way back in 2013, Diana Reiss, Neil Gershenfeld, the musician Peter Gabriel, and I got together to do a TED Talks about what would happen if nonhumans could use the internet.
Could they actually communicate with each other? Are they smart enough to have language? That’s always a loaded term, because normally we’ve associated language with humans. Yet, there is increasing indication that nonhumans do signal each other. I’ll use that word. We don’t yet know exactly what all the signals mean, however.
So these collaborative examples tell us that there is cognition in nonhuman species. It’s pretty obvious when you watch them solving a problem, or when you watch Peter Gabriel’s performance with a bonobo ape playing a piano and see [the ape] actually discover chords and octaves.
It’s clear that something’s going on there. So we’re very excited about the possibility of using the internet to allow nonhumans to interact with each other—but also to enable some of that interaction by using the internet and its assets to translate those interactions.
Lareina Yee: What are some of the capabilities or technology breakthroughs that we’ll need to enable that?
Vint Cerf: In some sense, the machine learning models take large amounts of information and find correlations. And sometimes, they find causative elements so that you can predict what some utterance might mean. We’re seeing a lot of that happening with machine learning models in other domains.
It’s not unreasonable to expect that it could discover correlations in the signaling domain. As I say, this is all still very speculative. But it might help us understand other species better. And it might change the way we treat them, which might turn out to be a good thing.
What kind of wine would the internet be?
Lareina Yee: On a more personal level, one thing you and I have talked about before is your passion for wine, and I believe you have quite an enormous collection.
Vint Cerf: I wouldn’t say enormous. Friends have very substantial collections, like 10,000, 30,000 bottles. I have maybe a couple thousand.
Lareina Yee: This is something I’ve asked you before, which is if the internet was a wine, what would it be?
Vint Cerf: Oh, boy, what an interesting question that is. Of course, I’d want to associate it with my favorite wines. On the white side, that would be white Burgundy. I don’t think there’s anything better than a beautiful Montrachet. So the internet would have this richness, and this wonderful, flinty, mineral-like flavor, and not overoaked, which is a problem with a lot of chardonnays.
Then, if you’re looking for really big depth and significant stuff, cabernet is hard to beat. Then there are these other little interesting offshoots that people don’t know about, like Tannat from Uruguay. Then you get to the Syrah, petite sirahs, or the Shiraz, as they call it in South Africa and Australia.
So if the internet were a wine, I think it would be a lot of different wines. It wouldn’t be just one.
Find success by embracing opportunity
Lareina Yee: It is many things. It’s also many possibilities. Thinking about the youngest generation and your time as a student, what are the things that you would suggest students today spend their time on?
Vint Cerf: Many students come in and say, “Can you help me plan my career?” And my first reaction is to explain to them that nobody who is successful has ever planned their career. It’s always been a surprise. The one thing that [these successful people] will tell you is that they were cognizant of opportunity when it showed up.
Sometimes opportunities show up in weird ways, like your company went bankrupt or you got fired. But it turns out to be an opportunity looking backwards, because you had to look elsewhere.
The second thing I tell kids is that if you want to do anything big, get help, especially from somebody smarter than you are. Bob Kahn and I were very good at getting help. That’s why the internet is what it is today.
The third thing is that if you really want a successful career, find something that you can do really well and that you like to do. Those are very simple, little rules.
Some students ask, “What should I study?” And I sometimes tell them, “You should study astrophysics.” They’ll say, “Why?” And I’ll say, “Here’s the situation. One hundred years ago, we thought we knew everything there was to know about the universe and we just needed to measure the constants more accurately to make better predictions.
“Then Einstein comes along and blows up the Newtonian universe. Then the quantum physics guys blow up Einstein. And now, we’ve got the string theory guys blowing up the quantum physicists. So, at this point, it turns out we don’t know very much about the universe at all. Which means if you go into astrophysics, the chance of winning the Nobel Prize is very high, because we don’t know anything.”
Lareina Yee: This is amazing. All the ways in which we can progress and advance far above where we are now. I love the fact that we’ve also talked about the human factor of technology.
