AIM Interviews

Leonard Kleinrock

Leonard Kleinrock. I am Distinguished Professor of Computer Science at UCLA and have been on the faculty since 1963. I developed the mathematical theory of data networks, the technology underpinning the Internet, while a graduate student at MIT in the period from 1960-1962. In that work, I also modeled the packetization of messages and solved for a key performance gain that packetization provides. Working with my extensive team of students, programmers, and staff, the birth of the Internet occurred in a UCLA laboratory (3420 Boelter Hall) when our Host computer became the first node of the Arpanet in September 1969 and it was from there that the transmission of the first Arpanet message was sent on October 29, 1969.

SK

In 1961, you wrote your thesis on queueing theory applied to message switching. This became the theory underlying packet switching, which enabled distributed networks like ARPANET and the modern internet. What originally made you interested in message switching? Was there any inspiration?

Leonard Kleinrock

As a PhD student at MIT in 1960, I was looking around for a challenging research problem which was unstudied and which would have impact. I was surrounded by powerful computers at MIT for which there was no technology by which they could communicate and I knew that sooner or later they would need that capability. Here was an unmined and important problem, whose solution would have impact, and I had an approach how to solve it. This was exactly what I was looking for. Using the tools of queueing theory, I constructed a mathematical model of message switching network technology which could properly support data communication between computers. I then analyzed, optimized and extracted the underlying principles of that technology. This provided the mathematical underpinnings for the packet switching technology of the Arpanet.

Leonard Kleinrock

Independent of my theoretical analytical work, Paul Baran at Rand Corporation was simultaneously studying the architecture of data networks that could provide protection against a nuclear attack, and some years later, Donald Davies at the British National Physical Laboratory was studying operational and implementation aspects of such networks. My research established the viability and huge efficiency of computer networking and this helped convince Larry Roberts, Chief Scientist at ARPA, to allocate the funding to create the Arpanet.

SK

You established the Network Measurement Center in 1968 and set up the first ARPANET node at your lab in UCLA, which later famously sent the historic first ARPANET message. At the time, what did you believe would happen for the world if ARPANET succeeded?

Leonard Kleinrock

Yes, because we had the mathematical strength and the software expertise, my laboratory at UCLA was chosen as the first Arpanet node whose function would be to serve as the Network Measurement Center. It was our job to anaylze network behavior and to evaluate, stress and uncover faults in the fledging Arpanet. On October 29, 1969, at 10:30 pm Pacific time, we sent the first Arpanet message between the first two Arpanet hosts (from UCLA to SRI). The message was simply "Lo" as in "Lo and Behold" (I added that later) and turned out to be a powerful, profound and prophetic message.

Leonard Kleinrock

Regarding my vision as to what would be the impact of this network, there is published documentation of my thoughts at that time. Indeed, months prior to the first Arpanet IMP installation, UCLA put out a press release on July 3, 1969 in which I articulated my vision as to what this network would become. Importantly, the release contains a direct quote from me in which I presented the following vision for the coming network: "As of now, computer networks are still in their infancy, but as they grow up and become more sophisticated, we will probably see the spread of 'computer utilities,' which, like present electric and telephone utilities, will service individual homes and offices across the country." I am pleasantly surprised at how the "computer utilities" comment anticipated the emergence of today's Web-based IP services, at how the "electric and telephone utilities" comment anticipated the ability to plug in anywhere to an always on and "invisible" network, and at how the "individual homes and offices" comment anticipated ubiquitous access. So, yes, I did foresee many aspects of today's Internet.

Leonard Kleinrock

On the other hand, neither I, nor anyone else at that time, envisioned a key ingredient of today's Internet; specifically, I clearly missed the fact that my 99-year-old mother (now deceased) would be on the Internet at the same time my granddaughter was. That is, I did not foresee the powerful community side (the social network side) of the Internet and its impact on every aspect of our society. It was not until email swept the network in 1972 did I realize the power of using the Internet to allow people to interact with people (i.e., social networks).

Leonard Kleinrock

Moreover, one of my predictions, namely, that the Internet would be "invisible", has yet to occur. Access to electricity is "invisible" in that there are no passwords, keyboards, screens, etc to access electricity – all you need to do is to plug into a socket in the wall and you get electricity! Today's smartphones and powerful laptops require too much fiddling around to gain access to the Internet. Once we deploy embedded technology into our physical environment and deploy personalized intelligent agents into the Internet, we will then be able to interact with the Internet using our human-type mechanisms (simple presence or voice, for example) and then we will begin to reach invisibility. At that point, I expect the Internet will evolve into a pervasive global nervous system.

SK

You've been a mentor to generations of researchers and programmers who went on to make pivotal contributions to the internet. You supervised Steve Crocker, Vint Cerf and Jon Postel on the ARPANET project and co-authored the hierarchical routing paper with Farouk Kamoun. What do you think was the most important trait you looked for in the students you worked with?

Leonard Kleinrock

Indeed, mentoring research students has been the wellspring of my life. It is a constant source of pleasure and of exciting trajectories of discovery. Without doubt, the most important trait I look for when I select students to work with is their passionate curiosity. I look for students who not only can solve a problem or figure out how things work, but also why they work that way. Why does the answer to the problem look so simple, is there an underlying principle expressing itself, can that answer be extended or generalized. Those are the traits that I seek in a student, and once they begin to work with me, it is my pleasure to teach them how to do research. It is always a thrill to work with them to constantly be surprised with them as we uncover deep and beautiful results. I have been extremely fortunate to have had a small army of brilliant PhD students who have been able to fulfill those objectives.

SK

Are there any people or projects you particularly admire today?

Leonard Kleinrock

In addition to the amazing networking technologies that my own students have explored, the current work in the exploding field of artificial intelligence is exciting and challenging. I have watched AI grow from its infancy to today's colossus. Following the pointers from Turing, it was in 1956 when John McCarthy coined the term "Artificial Intelligence" at a Dartmouth conference with invited giants such as Marvin Minsky and Claude Shannon, with all three of whom I had the pleasure of studying at MIT in the late 1950's and early 1960's! AI's early beginnings were exciting and focused on symbolic AI, but even in those early years, biological neuron studies led to perceptron models which grew into the neural networks and LLM's that dominate today's focus and excitement.

Leonard Kleinrock

Today we are seeing huge growth in the field along with deep concern (including concern from some of the major contributors to the field) about the direction and control of these powerful systems. Indeed, I can recall an early warning signal about dealing with these systems (whose underlying "intelligence" is, in many cases, not "explainable") when the Arthur Samuel's 1960's checker playing program was accidentally set to lose instead of win, and Samuel could not tell this was the case; imagine if one of today's LLM's was assigned a critical national task and accidentally was set to "lose"! Such concerns are top of mind for AI pioneers such as Geoffrey Hinton and Yoshua Bengio. So the power of AI is a particular technology I admire deeply, but is also one for which I have some concerns. My hope and belief is that if we are to "control" the possible dangerous behavior of tomorrow's AI technology, it will come by using AI to police itself.

SK

What's a message you have for the world?

Leonard Kleinrock