Exclusive: Internet pioneer Leonard Kleinrock on the great experiment we’re living through, and why the internet probably won’t break
Fifty years ago, Leonard Kleinrock helped create the internet. In an interview via FaceTime, he reflected on the experiment of a world living more and more inside the internet during lockdown, why the internet probably won’t break under the strain, and how young scholars can develop the next generation of connectedness.
A little over fifty years ago, the laboratory of Leonard Kleinrock at UCLA sent the first packet of data over the Internet. Kleinrock sat down last month to talk with ZDNet, via FaceTime call, about the phenomenon of aa life lived more and more on the internet amidst the COVID-19 quarantine. In his view, "the world is now conducting an experiment they never could have conducted on their own," the result of which is not just that we'll learn things about society, but that "we'll never return to where we were."
Kleinrock helped establish the theoretical underpinnings of the Internet in the early 1960s with his Ph.D. thesis for MIT,"Message delay in communication nets with storage.". Kleinrock's work was combined with research by two others, Paul Baran of the RAND Corporation, and Donald Davies of the UK's National Physical Laboratory when he was asked by D.A.R.P.A. to help design the Internet. (More on the history of the Internet is available in an exquisite article Kleinrock wrote for IEEE Communications Magazine in 2010.)
There is a "vanishingly small probability," Kleinrock told ZDNet, that the Internet would break under the current load. It's more likely the Internet will be just fine, he says.
Kleinrock is very active in technology today along numerous lines. In addition to doing fundamental research into blockchain technology, he is helping to develop an alternative social media platform, backed by investors The Sunday Group, known as Mobby, as ZDNet reported last October.
Kleinrock is very passionate about mentoring young scientists at UCLA, for which he has established the UCLA Connection Lab, a multi-disciplinary program that is intended to "support advanced research in technologies at the forefront of all things regarding connectivity." Kleinrock reflected on his own lessons about curiosity and learning as a young student at MIT. He urges scholars to spend time not merely simulating calculations but asking "what the numbers are telling you."
ZDNet:Congratulations, professor Kleinrock, now everyone is living in the network you built.
Leonard Kleinrock: I helped build, I helped build!
ZDNet:You helped build. What do you think about the fact everyone is in some sense living inside that network?
LK: The response is that it's enormously gratifying to have been able to do something which is serving so well at a time of crisis like this. Before this, it was a part of daily life as well. But here it's a really necessary component. And I think it's actually standing up very well. It's not crumbling, it's not getting that overloaded. What's interesting, I think, is that the load from homes typically was in the evenings before this virus, people streaming video. Now, it's spread out over the day. And it takes advantage of the enormous bandwidth that we have personally from our homes that was lying idle during the day.
The thing that I recognize -- I'm sure others have as well -- the world is now conducting an experiment they never could have conducted on their own. You know, everything shut down, people working at home. So, on the positive side, we have an opportunity to observe this change, make measurements, experiment in a way we never could have before. I imagine there are Ph.D. students all over the world taking advantage of this experiment they didn't set out to create, and observing from what's going on, so that's great.
But the flip side of that is, as I'm sure you're aware, we'll never return to where we were. Too much has been experimented with, which has proven good and bad. I believe the recognition that online education, remote education, works in some environments, in some cases, is now clearly demonstrated whereas before it got hypothesized it would never work. It's working, in universities, in elementary schools. It's also challenging at a deeper level the value of a campus-based college education, the economic issue, the value issue, how much college education is social interaction versus putting stuff in your head. And the answer is not clear, but the question is there now. And I think parents and students are going to be asking that question. So I think it's going to be a sea change in terms of the economics, and the structure of higher education, the fact that the faculty are learning how to do online where they never would have done it before.
Some mix of work at home will undoubtedly remain. And some jobs that used to be necessary are now being seen to be not necessary. Businesses are saying, "Gee, I didn't need that functionality anyway. We can get it either by AI or by some other automated way". And entertainment, you know, do I really want to go to the movie theater? Well… versus Netflix or whatever? We're never going to get back to where we were.
In engineering, there's a term called hysteresis. We're in such a situation now, where we've stretched the system in one direction. If we relax now, it's not going to come back to where it was. It's going to have memory of what went on. It certainly applies to medicine, to social interactions, etc. So, I find that very exciting. It is exploring things we couldn't have, and we're finding advantages of those things. Economically, it's a very serious issue here, what's happened and how we come back. Supply chains are being broken. How they get restructured is not clear.
There are opportunities out there now for new products and services based on the fact that we are less in physical contact, more remote. There are services that will arise that we haven't seen before involving how we interact both socially and physically and shopping-wise. I think of a few of them myself, a few possible products. But, there's a world of opportunity now to match the needs of the world we are living in now and will be living in.
ZDNet: Do you care to offer a couple of those product opportunities that you're thinking of?
LK: If you look at what's happening to currency, dollars, and suchlike, we're in the world now of fiat currencies. We're moving toward a world of blockchain, of virtual currency. Moreover, we're also in a world of other virtual currencies that exist, for example, in gaming worlds. I expect there's going to be a kind of virtual currency that gains traction because of the virtuality of the world we live in now. And what it's going to do to the fiat currency is not clear. I've been studying blockchain for quite a while now. Will there arise a way in which it's a kind-of glorified barter system? And what will be the currency that dominates it? You can imagine that a psychotherapist can offer services over the 'Net, perhaps, and get in return for that training on a computer. Now there's no currency there, it's a service and a service. But maybe there needs to be an intermediary currency of some sort. The huge disadvantage of a barter system is that pairwise needs must find each other in order for the transaction to take place; with currency (gold/silver-based, fiat or virtual), that disadvantage disappears and that was one of the drivers that allowed "money" to gain such prominence. With the Internet, there are effective ways to find matches, so that obstacle is less of a challenge. That means that now we have the opportunity to explore the role of virtual currency and other features of blockchain distributed ledger technology.
ZDNet: A store of value and/or, a transaction system, maybe both…
LK: Right, exactly. So I'm just toying with the idea. I just sense there's something that's going to come out of this. It's not going to remain as is. And blockchain is driving a lot of that, of course.
ZDNet: But it's not Bitcoin, or Ether, in your view?
LK: It's not clear. It certainly is a good candidate. And I'm very much in favor of blockchain in its various forms. But I think services will be a part of it, as opposed to only being a virtual currency or a hard currency. I don't know. It's a very early thought, but -- there's something there and I sense there's a movement here that's going to really take hold.
ZDNet:Looking back to the early days of the Internet's development, did you think, way back then, that someday there would be what you're describing as this kind of impromptu experiment where something dramatically changes, and suddenly all the remnant bandwidth is being used? I mean, it would have sounded crazy in 1965 or something to describe this, right?
LK: The short answer is, No, I did not think such an experiment would be "run". But, there was an underlying recognition and a drive that helped lead us to where we are, and it's the following. For me, the creation of computer networks was a new, challenging, unstudied problem whose solution would have impact and for which I had an approach. As a student of Claude Shannon, I realized that if you build large systems, with large numbers of participants or large numbers of nodes in the graph, then there are emergent properties that appear that you won't see in a small network. My thesis proposal [for the MIT Ph.D.] was titled, Information Flow in Large Communication Nets, the emphasis on large. Now, if you're going to build a large system, in the case of a network, you can't allocate all the control to a single node, for many reasons – you gotta distribute it. And if you distribute control so that it's not vulnerable, you know, you can chop off a piece of the network, the rest of it continues to flow because the control is all over the place. If you do that, you automatically get robustness as a bonus to the need for scalability. It wasn't a driving criterion that it should be robust and could withstand an attack or a failure. That came automatically, for free.
Now, if you have a scalable network, a large network, you're going to have lots of capacity out there. And depending upon how the traffic flows, it'll need more or less, but it'll be plenty of extra capacity. So, extrapolating that to today, we built this network which had capacity where you needed it, and where you might need it. The other part is, before the pandemic, we had this night-time heavy use of the network from homes. But the capacity was there all day long and you could deploy it to be used in a dynamic fashion on an as-needed basis. Well, we need it now, and we're using it all day long now. So in a certain sense, the scalability in the large network allowed us to build a network that would have this property, but we didn't anticipate the need for what we're seeing today.
ZDNet:What happens, then, when you take that remnant capacity, and you move around the utilization into different new, uncharted time points? Is anything interesting happening, technologically, in this case?
LK: Yes there are interesting things that happen. Are there phenomena we can identify? Not necessarily. There are strange things that happen. It's possible for a network to start sending flow around, and then if you add flow, but the preferred paths are already being used, so you take a less optimal path, and then you're boxing up somebody, you're using somebody else's optimal path, they take a less-optimal path, and you can get into a situation which is far from optimal. So you need this adaptivity, you need to be able to constantly sense and change and test. And we have that in most of our routing networks today. You've got to be careful you don't make it too optimized or too rigid. You have to allow it to adapt at all times. Otherwise, you're likely to run into a situation you did not anticipate, and it's going to crash.
ZDNet: And the question that the journalist asks at this point is, are you telling me this thing's going to break?
LK: Yes, that's possible. Because in those early days of testing --UCLA was the network measurement center -- it was our job to try to break the network, to find faults, so we could fix those faults. And we could constantly break it. We'd find protocol, hardware, software failures, and they got fixed. And in those early days, we found them all the time. But those failures were not obvious until they happened. And the problem is that the way in which the network was controlled was, uses a term called flow control, where you put constraints on the system. I'll give you an example. A flow control constraint is anything that affects traffic flow between the input and the output of the network…. You say, for example, you may want to say, the order in which data comes into the network is the same order in which it should come out.
The point is, any time you place a constraint called flow control on a network, you're opening yourself up for failure. If the network can't meet your constraint, then it is likely to cause a failure. If it's slow in meeting the constraint, then you may get a performance hit. Now, sadly, the way the Internet has been designed, there are flow controls, constraints all over the place. You know, somebody adds a new application or a service or networking functionality, and they put a controller in it. And it's done in a distributed fashion, in a way that's -- this is a negative now -- what you add as a flow control, and what I add as a flow control, we don't necessarily know about each other. And it could be conflicting or problematical, and we can't get our hands around it. So, in answer to your question, will the network crash, I'm convinced there are situations, if the network ever found itself in those conditions, that would activate a disastrous flow control interaction that could take the network down.
ZDNet: Can you give an example?
LK: If I could, I would fix it!
ZDNet: So, the next obvious question is, Is the Internet under this current unprecedented situation going to break given know unknowns? Can you say that?
LK: There is a probability. It's a vanishingly small probability, but you can't guarantee that it's zero. There are constraints on the network which could cause catastrophes if you ever reach a situation where you activate those constraints. And that's where the unlikely part is. I mean, we're running a network for 50 years now. We found some problems, fixed them as we found them.
ZDNet:I guess the better way to put this is, there is an overwhelming chance the Internet will be just fine in this situation.
ZDNet: You have over the years talked about some fascinating ideas about user interaction, which you call the Invisible Internet, and also Smart Spaces. I am, perhaps unreasonably, associating this in my mind with touch-less interfaces. It seems like germaphobes would be crazy about something like Smart Spaces now.
LK: Well, it's a very good point. I like that. The reason I call it The Invisible Internet is the following. Back in '69, before the network even started, before we got that first switch, I said in my press release at the time that we'll see that it be available all over the country, from homes and offices, blah, blah, blah, and it'll be like electricity because electricity is invisible. It's a beautiful interface. Two holes in the wall. It's a simple interface, you plug in and you get electricity. It's about as invisible and easy to use as you can imagine. And in those days, I argued yes, the network should look like that, should be as invisible as that. Now, the Internet right now is anything but invisible. You've got a complicated interface on this phone. All the keyboards are different, the applications are different, the various functionalities are different. It should be that when I walk into a room, the room should know I'm there. They should know it's me. And I should be able to talk to the room in my native language, and get a response in my native language on a display that pops up, maybe on a hologram, maybe with some gestures and haptics, et cetera, in a natural way. It should be as invisible and everywhere as that. It's not there yet, but we're moving in that direction.
And the Internet of Things is helping us a lot because we're beginning to deploy embedded technology in our walls, in our desks, in our fingernails, in our bodies, in our offices, in our cars, which have actuators, sensors, displays, microphones, speakers, cameras, logic, memory -- all that functionality, so that I should be able to interact with it naturally. I mean, Siri is a little bit of an example of that. And it's beginning to happen, but I believe it's going to allow interactions by the common man far more easily than it is now. And allow people to have a voice, with some protection, privacy on the one hand, and a limit on how loud a megaphone you get depending upon what's being said.
The great thing about the Internet was, it gave every person, no matter how poor or dirty or small, with a laptop and Internet access, a voice where they could reach millions of people immediately at no cost, no effort, anonymously.
Well, that's also a perfect formula for the dark side. You put a malicious person there and they abuse it. I believe there are ways to protect against a lot of that if we give the appropriate ability to speak up, the ability to not be inundated with fake news, so we need some detectors of that. I think, A, the invisibility, and B, a thing like a distributed ledger where everything is on there and can be investigated and evaluated, will help give us this world we hoped to provide originally at the birth of the Internet, which is free, ethical, shared, open, trusted, and invisible.
ZDNet:So, what happens to Smart Spaces in an event like this, does it get new momentum or does it take a back seat because just basic connectivity is the priority?
LK: That's an interesting question. I think it's going to continue to be deployed because we're going to spend time in less-endowed environments. If we spend less time at an office, which is full of high-speed printers and good machines, we're going to want this functionality to be available in Central Park. Back in the early '70s, I would show a picture I got from Datamation magazine of a woman sitting in a chair, in the middle of Central Park, at a portable terminal, communicating, no wires, no power, and no wired communication, all wireless, back then, in the '70s. And so, the IoT and everywhere access were there.
ZDNet: So you're saying given that IoT proliferated in relatively benign times for the electronics industry, maybe it gets a continued push during this period of unprecedented kind of duress?
LK: Duress, and non-endowed environments. We're gonna be out in the street. Our car can have it. Your office can have it. But can your home, can your bicycle, etc.? Well, we're beginning to deploy these things. But the problem with IoT, the big problem, is privacy. And the abuse of that. Even Siri is an abuse if you think about it. And Alexa and all the rest. And right now this camera, and this conversation we're having, it could be tapped and it could be being sent to the Kremlin or Beijing right now.
ZDNet:Mobby is one approach to try and address some of that. Are there other kinds of things that need to happen?
LK: Empowering people, I think, is starting to happen, the tools are not there, but once they can interact in a more trusted fashion… Trust is critical. And the blockchain world does spend a lot of time talking about trust on the one hand, and survival against malicious users on the other hand. Related to trust is the notion of reputation and we are researching how to incorporate reputation as a first-class entity in the Invisible Internet.
ZDNet: Blockchain is one of the things you're working on at the Connection Lab at UCLA?
LK: Yes. The Connection Lab is a hotbed of activity not only in blockchain but all things connectivity-wise. The UCLA Connection Lab is a newly funded research center devoted to shaping the future of the internet and computer networking. It is an environment of collaborative, interdisciplinary, open research, with an underlying theme, and that theme is connectivity.
ZDNet: What's the broader vision for the Connection Lab?
LK: The UCLA Connection Lab is designed to provide an environment that will support advanced research in technologies at the forefront of all things regarding connectivity and will deliver the benefits from that research to society globally. The Lab's broad-based agenda will enable faculty, students, and visitors to pursue research challenges of their own choosing, without externally imposed constraints on scope or risk. It will draw inspiration from UCLA's foundational role as the birthplace of the internet.
My goal is to make it a highly interdisciplinary, collaborative environment. Not only with graduate students and professors, but with undergraduate students from across campus, in the humanities and business, and social studies. In medicine. Because, if you focus only on, engineering, you're going to get too narrow
The UCLA Connection Lab is funded with donations from individuals and groups committed to the advancement of the use and understanding of advanced networking in society. Our inaugural donor is the Sunday Group, which has generously provided funds for creation, operation, and endowment of the UCLA Connection Lab.
Right now it's closed, obviously, because of the pandemic. We can't go there. When you come, if you come out to California when this quarantine is up, we'll show you – it's a beautifully designed collaborative research Lab.
ZDNet:And so is anything changed with all of this in the present moment? How has this affected that process?
LK: Well, we haven't fully populated this new Lab with people. My graduate students work there. We are working on blockchain technology, reputation systems, network control, IoT, AI, cryptography, security, and all things related to connectivity. We're bringing in more graduate students. We have invited the UCLA blockchain people to use it, to spend their time there, with other faculty, with their students. Right now, it's closed due to the pandemic, nobody's allowed on campus right now.
ZDNet: Sure, but everyone can still do some work remotely?
LK: Oh, they all are. I'm with my students all the time on Zoom as I am with you now. But the Lab's not fully populated yet because it just really opened up, some of the renovations were being finished as late as January.
ZDNet: That's to be expected. What happens this fall?
LK: Well, there's some question as to whether or not they're going to open up the campus for the fall. For this spring, no, everything is online. It'll open slowly, and once it's safe in a real sense to go back there, we'll begin to get students back. I'm sure the students will jump in very quickly. I want to be careful how quickly we move back because I'm worried about them. But they're doing very well at home, I must admit. But they're lonely.
ZDNet:On the subject of scholarship, you were a student of Claude Shannon, a towering figure. What qualities stand out to you about him?
LK: Great guy. What Shannon did was brilliant. He looked upon information as to how much uncertainty is there in the information you're sending. He never talked about the value of the information or the application. So his theory was application-independent, which made it very general. Occasionally he was criticized for not giving pure mathematical proofs that the mathematicians considered solid. In some cases, they had a point. But, he was always right. His brilliance was unmatchable.
ZDNet:Did you feel like, over time, you came to appreciate what made him tick?
LK: Yes. Oh, it totally affected the way I view research. You'd walk into this man's office, he was a great mathematician. And in some sense a great engineer. What would he be doing? He'd have a differential gear in one hand, and a Swiss Army knife in the other, unscrewing the gear. He was very practical. He wanted to know how things work and got his hands dirty. So he had intuition, not only about mechanical but about mathematical things. So he could bring physical intuition to the mathematical precision. And that combination is extremely valuable. And in fact it made him think, if I got this mathematical result, that's nice. That's correct. I've proven it. What does it mean? What is it telling me? How can I extend it? What's the lesson here? Is there a principle being exposed? Now, that's the way I do my research. I ask those questions. And I grew up as a tinkerer working with radios, electronics, model airplanes, and solving puzzles – all intuition builders that trained me to look beyond the obvious.
LK: Yes. Until you get an intuition that you can then use. I have students come in, and I give them systems to evaluate. And nowadays, with the presence of very sophisticated computers, they can too easily simulate the system and come up with the numerical performance, the throughput versus the load. So I say wait a minute, now, why does this portion look like a straight line? What does the slope mean to you? What about the intercept where two lines cross? What's meaningful about that? What's the asymptote? Why did it start here? They say, "I simulated it". I say, what if I doubled the throughput somewhere – what then will be the result? "Well, I'll simulate it again". I say, you failed! I mean, because they have no understanding.
ZDNet:What is that? What is that ability?
LK: Curiosity, and intuition, and not just getting the answer, not just scoring on the exam. But rather understanding what the numbers are telling you.
ZDNet:And those are the kinds of things that can flower with the Connection Lab, it sounds like.
LK: Theidea of having other students around in different specialties, interacting with you, is the real benefit. I remember I had that experience at MIT. as a graduate student. We had all of the PhDs in kind-of a bullpen. Now, it was all electrical engineering at the time, but there are so many varieties of doubly-EE. And what happened is, my classmates, many of them were foreigners, in those days we just took the best from everywhere. I remember two of my classmates, an Israeli named Jacob Ziv, and an Indian named Tom Kailath, both very successful now, they were in totally different research domains. And I remember they would go to a blackboard, and they would teach each other their specialty. Just one on one, their specialty and also their research, and they'd critique each other. That was a wonderful environment just rubbing elbows, learning. You don't get that in this environment now, you see.
ZDNet: You don't quite get the same experience in Zoom in the same impromptu way.
LK: No, you don't. You don't. Although would I be looking at you this way? You know, we'd probably have a telephone call otherwise… It's good to meet you this way though.
ZDNet: It's good to meet you too, professor! Thank you for taking the time!