E44: Peter Lee, Corporate VP, Head of Microsoft Research – Interview

July 28, 2016


This interview with Peter Lee was quite a treat for me. Peter is the Corporate Vice President, Head of Microsoft Research. I know it’s not all fun and games but sounds like a pretty amazing position to me.

Before that Peter was a long time professor at Carnegie Mellon where his research included cybersecurity and software reliability. In 2016 Peter was also appointed by President Obama, as a member of the cybersecurity commission.

In 2004, he took a leave of absence to go to DARPA, where he founded and directed a new technology office that created research and development programs in computing and related areas in the social and physical sciences.

Here are other topics we talk about:

-Why did you go to DARPA? How was that experience? Tell us about the red balloon competition.
-What are your main priorities as VP at Microsoft Research?
-What’s one technology you’re especially excited about?
-How is working for Microsoft different than Academia?


David Kruse: Hey everyone. Welcome to another episode of Flyover Labs. This is Dave Kruse from Madison, Wisconsin and today we have renowned researcher with us Peter Lee. And Peter is the Corporate Vice President and Head of Microsoft Research, and before that Peter was a long time professor at Carnegie Mellon, where his research included cyber security and software reliability. And in 2014 Peter was also appointed by President Obama, that President, it’s like the President of Presidents as a member of the Cyber security Commission and in 2004 he took a leave of absence from Carnegie Mellon to go to DARPA where we Founded and Directed a new technology office that created research and developing programs in computing and related areas in the social and physical sciences, which sounds quite interesting. So I’m excited to hear – I brought Peter on, so I’m really excited to hear about this background and the research focus and why he switched to Microsoft and talk about, more about the Cyber security Commission if we can. So Peter, I really appreciate you coming on the show, thanks for joining us today.

Peter Lee: Thanks David. It’s great to be here and thanks for having me.

David Kruse: Definitely. And so before we jump into our research now, can you give us a little background, like where are you from? Were you always interested in technology or when did you start getting interested? Just curious to learn a little bit more about you?

Peter Lee: Yes, sure. I actually grew up not too far from Madison, in upper Michigan a little town called Houghton. So I guess that would be life right under, country.

David Kruse: That’s right.

Peter Lee: And I had, I’m Korean inherited. I was born in Columbus but my parents emigrated from Korea and my dad was a physics profession and my mom a chemistry professor. So pretty – you can imagine pretty hard core Asian science household, and then went to University of Michigan, probably disappointed my parents to study math instead of physics or chemistry.

David Kruse: That’s so easy, the math.

Peter Lee: And, then eventually got very excited in computer science and finished my PhD. Went to grad school at Michigan in computer science and then from there, after my PhD went to Carnegie Mellon to be a professor. And I was there for a long time, for 22 years and untimely the Head of Computer Science there, and then as you mentioned I went to after – in 2008 actually I went to DARPA and really learnt a lot there and then from there instead of returning to Carnegie Mellon, ended up here at Microsoft.

David Kruse: Got you. And so what year did you start at Carnegie Mellon?

Peter Lee: Way back in 1987 and yes.

David Kruse: Was your research focused around security at that point?

Peter Lee: You know back then I was just out of grad school and my PhD research was really in a much more theoretical part of computer science, which was called formal semantics and that’s, I think most people consider it a very theoretical, really abstruse part of the field of computer science where you are trying to use mathematics to understand the meaning of programs and programming languages. And when I got to Carnegie Mellon, one of the things I got exposed to was what we now call cyber security, computer security and one of the challenges in computer security of course has to do with keeping secrets, which is handled by different forms or inception or cryptography. But another big part and even bigger part in cyber security had to do with how do we know what a program will do. So if I were to send you a piece of code and ask you to execute it, how can you trust that it’s safe to do that, and that caution has become a bigger and bigger one over the past 30 years as the internet has sort of become a platform for code sharing. And so all of the theoretical work I had done as a PhD student in trying to kind of mathematically or formally understand the meaning of programs, it ended up being very relevant to computer security.

David Kruse: Interesting, that makes sense. And yes, you see that a lot where you kind of combine two areas of expertise and you become like the expert, that’s interesting. And so are you still working in – because you are right, like with the interest how do you know if a program or application is safe to open or what’s it going to do exactly. Are you still working on that space and how – and if not or how has that space evolved since you started researching it back in 1987.

Peter Lee: Right, well you know in my position as a VP here now in Microsoft Research, of course I still have a real interest there, but one thing about age and as you move up the corporate ladder is you tend to get less and less technical or more and more stupid, I don’t know which one.

David Kruse: Fair enough, right.

Peter Lee: And so basically management issues really end up consuming a lot more of my thought process, but on top of that you know Microsoft research is a pretty big place. I think we have about 1,200 researchers in 11 labs around the world and so we cover a lot of space and so for sure computer security is a big area and generally the trustworthiness of the Internet and of mobile phones and of laptops and the Windows operating system, all of that, that’s a big deal. But there are other areas like machine learning and artificial intelligence or the architecture, the computer systems architecture and network architecture of our cloud or you know the kind of room tracking and computer vision capabilities in optics, in hollow lens. All of those things end up being areas where I have some management responsibility now at Microsoft Research. So you know my heart will always be in computer security, and the people who work in computer security in the labs here, maybe they get a little bit of extra special attention from me, just because I can pretend more like I know what I am talking about.

David Kruse: Want or unwanted attention.

Peter Lee: Right, but you know it – but in reality day to day you know I have much broader kind of sort of topics now to worry about.

David Kruse: Definitely. And I want to talk more about DARPA, but first I think you probably have one of the most interesting jobs in the world in my opinion, because you do get to touch on soo many interesting things, probably with some of the best experts in the world. Yes, so what led you to Microsoft? Did you go to Microsoft as a Head of Microsoft Research or what was your initial…

Peter Lee: Right. No, I started a little bit for like – gee, you know all of these things feel accidental to me honestly. I you know – when I reached the point of being Head of Computer Science to Carnegie Mellon, I really thought that I had the best academic job in the world and it’s such a wonderful place and it’s like just a stupendously great computer science department and amazing stuff happens there. Do you know in 2008 when President Obama was elected, a friend of mine, a guy named Tom Kalil got appointed the Policy Director for Science and Technology Policy in the Whitehouse and so you know he leaned on all of his friends to write little two paged position papers on what the new administration should be doing in science and technology and I foolishly agreed to write a short two pager about DARPA and so – and I have some knowledge about DARPA because I was on various advisory board and so on for that agency, but the real impact of writing that wasn’t so much the ideas I put in the white paper, but that was used to pressure me into going to DARPA to serve. And so – but then I had a wonderful time at DARPA and I really grew both intellectually and just as a leader a lot there, and while I was there, I would run into other people at different companies who also would have an interest in someone who was willing to work in kind of management positions in technology or tech innovation or research and a couple of those people were Rick Rashid who was the Founder of Microsoft Research and Craig Mundy, who was essentially the Chief Technology Officer for Microsoft at the time. And they slowly and steadily worked me over and exposed me to the idea of coming to Microsoft Research and so when I finally did – agreed to do that, I came as the Director of the Microsoft Research lab in the headquarters here in Redmond, Washington and so that’s how I got my start here.

David Kruse: Got you, okay. And at DARPA, you know you mentioned that you grew a lot. So how did you grow and what did you learn that was different than at Carnegie Mellon.

Peter Lee: Yes, DARPA is really a magical place. I don’t know how much your listeners know about this. It is – this name has changed between ARPA and DARPA and ARPA stands for the Advanced Research Projects Agency and then if it’s called DARPA, it’s the Defense Advanced Research Projects Agency and DARPA was created in the year after the Russians launched Sputnik and so that was sort of a carthadic experience for the event, for the US kind of National Defense Establishment, because it suddenly opened up space a possible frontier for our National Security and for military operations and the Russians were ahead and that surprise really shocked the US Defense Department and so DARPA was created a year after that happened, with a mission to prevent technological surprise. It really did some amazing things you know. It worked on the booster rocket technology for the Saturn V that put men on the moon. They had invented the ARPANET, which eventually became the Internet. It invented the field of material science, did a lot of work on robotics, stealth fighters, drones. It’s just a pretty increasable place with amazing history. And maybe the thing that was kind of most amazing is its very failure tolerant. DARPA is willing to try lots of different things and most of the things that it tries end up not panning out and some of them end up being finally ridiculous in retrospect. But the one out of 20 or one of 50 times when a wacky idea ends up being working out somehow, it ends up chaining the world. There is a famous story in DARPA when DARPA had invented the idea of GPS and had been involved in developing the first satellites for the GPS constellation in space, but then there is question of these bulky GPS receives that had to be carried on ships or trucks. And so someone had the idea, well, maybe we can make them smaller, and small enough for a soldier to carrier and they got a bunch of top scientist together and concluded that it was physically impossible. The physics just wouldn’t support a very small device, but the legend in DARPA is that the Director said, I don’t care, that’s your mission, go execute. And that’s a kind of decision that I saw play out often at DARPA and often times the physicist is right, you end up failing because some things just aren’t possible. But sometimes the physics is wrong and we end up having GPS in virtually every pocket around the world and so that kind of thing, getting exposed to that and try to understand it and have it shaped how I think about managing research and tech innovation, it was just a tremendous learning experience for me and just the idea of tolerating sometimes some embarrassing failures, it was another growth experience.

David Kruse: That’s interesting, right and pushing the boundaries of technology, it’s probably easier to take a step back, oh it probably isn’t realistic, but to push it a little bit harder, you might get that result that can change the world, which is a…

Peter Lee: Yes, the very first project I had when I ran into DARPA, I had a bunch of what are called service chief fellows. These are kind of up and coming, kind of rising stars in the arms services, like that kind of Capitan level. And there is a program where a few of them get to spend three months with a DARPA office records. So I get these 10 service men and women and we have to – we are supposed to have a project. And three months project and so we do all this brain storming and there are all these ideas floating around and then we comp up this great idea to have a balloon hunt. We are going to have this big eight foot red weather balloons in these public but undisclosed locations, for eight hours on the 48 anniversary of the birth of the ARPANET and whoever can kind of those balloons, those 10 balloons first wins $40,000. And so we are just all charged up and excited about this. This is going to really test the power of social networks and so on. The day ends, I go back to my apartment and wake up in the morning and realize oh my God! What a crazy idea. I have no credibility in the US military and I’m going to spend $40,000 plus of our tax payers dollars on a wacky balloon hunt. There is just way I can do this, and so I was supposed to report to the DARPA Director that morning what our plan was and so I go to her office and I tell her, look, we really brainstormed all week. I thought we had a good idea, but I realized this morning it’s not going to work out, so we are going to need more time and so the Director said, okay, well tell me anyway what your idea was yesterday? And so I tell her the whole idea and she kind of thinks for a minute and she says yes, that does sound like a completely stupid idea, but this is DAPRA and we know that disruptive ideas are fragile and fleeting, that’s the phrase, fragile and fleeting. That’s the idea you through was good yesterday, that’s what you are going to execute. And so she forced me to do it, and it turned out to be a very important experiment and demonstration. But that kind of culture to kind of really put yourself into very uncomfortable spots was – really DARPA was the first place I got exposed to that.

David Kruse: Yes, what a wonderful environment and that sounds exciting. Can you give the results of the balloon hunt? Just for people who didn’t know. I remember it clearly. It was – I through it was a pretty sweet idea, but yes.

Peter Lee: Yes, it was an amazing thing. There were about less than 400 teams that registered for the competition and research teams from all over the world. I mean we had entries from China, from Russia, even a couple of entrants from Syria. It was war and so there was a Russian hacker group that hacked the website that had the, like the FAQ for the competition and they put a virus there so that if you are using certain web browsers, if you went to that webpage you could get infected with a virus that when you went to report balloon sittings you would actually get redirected to a spoof website run by that Russian group.

David Kruse: No way, that’s clever.

Peter Lee: So it was a wild, wild thing. In the end a team from the MIT Media lab ended up using various social media and data mining techniques to find the location of all 10 balloons in just under nine hours and by the air it was 56 minutes, which is truly an amazing accomplishment. And there are quite a few other teams that in that same time period had found either eight or nine of the balloons. And so there were – it was an amazing thing and these are all teams without any of the resources of spy satellites or armies to direct. They were just using social incentives on the internet to kind of recruit people and you know sift through all of the chatter. And it really kind of proved a point and the ideas that came out of that ended up being impactful in our operations in Afghanistan. But also other organizations like the United Nations and various political campaigns in different parts of the world ended up using social, some of the social structures and social media incentives that were kind of developed by various teams in the DARPA network challenge in their own campaigns. So it was really a very, very interesting and kind of high impact demonstration after all. So I’m so glad I was forced to do it.

David Kruse: Yes, that’s right. A crazy idea that turned out to be a brilliant. And we haven’t even talked about Microsoft yet and I know we only have so much time, but I am curious. I’ve always been curious of this. How do you decide where you put the balloons?

Peter Lee: That was – there was a lot of secrecy there, and so we had a, there is a standard sort of concept of operations for that kind of thing that the 10 service men and women that were interning with me had developed. There were actually 12 balloons, as there were two backup locations in case anything went wrong. And we had do very secretive things, because there were teams that are actively monitoring sales of helium around the country.

David Kruse: What.

Peter Lee: So it ended up being actually quite an operation with real military con-ups involved.

David Kruse: That’s cool, all right. Interesting, and so how did you decide where to locate them?

Peter Lee: So we wanted a – first of all we wanted a diversity of spots and we also wanted things to be very public and so in areas where anyone in public could be allowed to go. And so we did – first of all broke things down geographically into three segments, Eastern, Seaboard, West of the Rockery’s and central US and then we divided things between the urban, the suburban and rural areas and then finally we wanted areas that had a mix. We did have a navy operation that was going to use some satellite systems. So we also just wanted to test in different kinds of areas and kind of what [Technical Difficulty] balloons I think that was reported. It was a balloon that was in Union Square in San Francisco and that was reported by a team of psychics.

David Kruse: No way.

Peter Lee: That was the only balloons that they could see I guess using their physic powers that ended up being correct.

David Kruse: You never know.

Peter Lee: Right.

David Kruse: We only have a few minutes left, so let’s move on to Microsoft and I mean like I said, you probably have one of the most interesting jobs in the world. I mean I’m sure, day to day it’s going to feel like that, but man what – you have so many different interesting areas, each one which could be a podcast on its own. What are some areas in technologies that you are especially excited about that you kind of have an overarching feel to them, like across all Microsoft or maybe just – it just doesn’t have to be an overarching feel, but just something that you are excited about.

Peter Lee: Sure. Well I think you know artificial inelegance and in particular machine learning is really just so huge right now. You know it sort of started I think for us, with the – dealing with language, with human language and a huge amount of machine learning research went in initially into to the big search engine, because you are trying to understand the world wide web and all the web pages, but also all the queries and links that are picked on and entered by users. But as our capabilities and research has advanced, we are just getting so close to being able to have the ability to recognize speech, human speech, to translate languages, to understand what people are talking about. So the whole area of language processing is just leading to lots of things. For example we worked very hard to put automatic language translation into Skype, instead of using Skype at windows, that’s now a built in feature at least for eight languages, and you know it’s not perfect yet, but the things about the modern day eye systems is that they get better with experience. These systems are constantly learning and so the more they are used, the better they get and so that’s just clearly excited and that’s now moving beyond language to vision, computer vision and there is just so much possibility there as well. So AI is a big deal. Another one is in computer architecture and systems. There are could measures growing, about doubling every year and its huge. There are like a 130 data centers around the world now. But we are starting to reach within the few years, the end of Morris Law [ph] and so exactly how we maintain the growth to meet demand in cloud computing is leading to lots and lots of interesting new kind of computer architecture and systems research questions. And so those two things end up being really a major kind of areas that we are thinking on a longer term.

David Kruse: Got you and…

Peter Lee: You know you said a couple of times about having a fun job. It is fun, there is a lot of pressure. The way I sort of describe the job here, not just for me, but a bunch of my colleagues is it’s a lot of like the first time riding a rollercoaster, its – I remember as a kid the first time I road a rollercoaster I was so jazzed up and excited and I would get strapped in, and then as I went up the first hill slowly I realized that it was big mistake, I needed to get off and then by the time you finish that first ride, you want to ride it again. And I think right now in the tech industry and in my job it’s a lot like that. It is fun and exciting, but it’s also scary, because the pressure to innovated and invent is getting more and more extreme, not just at Microsoft but I think across the tech industry.

David Kruse: Definitely, and right and then otherwise, I mean on an annual basis you kind of have to show steps towards new products, new innovations and what have you seen is the best way to innovate to kind of take the giant leaps as far as like size of the team, where ideas come from. What has been your experience?

Peter Lee: Well you know, in my experience the smartest people thrive if you gather enough of them together, so it’s a critical mass and then leave them alone, make sure they are well resourced and leave them along. Any attempt to kind of meddle ends up being kind of productive and so then the trick is when ideas start to mature to the point that we could bring them into products and or into practice, you then need to somehow create the right kind of team with lots of engineering and kind of product design support to harness a good research that’s been developed and then really kind of have almost kind a skunk works kind of level of like sort of mission focus. And so one of the things I thinks about the most is exactly when is the right time to kind of harness some research knowledge and otherwise you try hard to resist the temptation to kind of meddle in the research that people are trying to do on a day-to-day basis and I get that wrong a lot anyone with that, you know it’s sort of fundamentally the – innovation is such a weird intangible abstract things. Roughly speaking everyone including me is doesn’t sort of by [inaudible] pants. But I think of it as a harvesting model. You are allowing great ideas to kind of be incubated and nurtured and then you are trying to find exactly the right time when to go all in on something and then try to form research ideas into – and results into some great new experience or great new solution that has some real value.

David Kruse: Definitely and like you said timings around that is well, there is no exact answer. So it’s much a qualitative as quantitative trying to figure what that is and I know we are pretty much out of time, but one last question was, how do you come up with new ideas? I’m sure coming up with new ideas isn’t always an issue. But how do you come up with ideas and then invent them? Because you probably have so many things you could be working on. How do – get from idea to where it is now, because that’s quite an innovative unusual idea.

Peter Lee: Well so, yes so I think about this in sort of three buckets and then the portfolio I try to have some balance in these three buckets. One is not everything is pure invention. A company like Microsoft has gigantic technical, hard technical problems that require research to solve and so one bucket of projects and investments that we make in researches just to partner with different teams and groups across Microdot to help them tackle their hard problems, so that’s one. But then there is a second one which is attempting to be more inventive or disruptive and you know for that part I have a team that’s a lot of like the managing partners and the venture capital firm and we try to encourage as much openness as possible and do things to generate idea flow and encourage people to come to us and pitch their ideas. And we rarely say yes or no when they come to pitch, but usually the response is, you have a nugget of a great idea. Once you work with us and lets keeping working on those ideas, to get it to a point where it’s interesting. And then the third bucket is a lot of focus on alternative markets and especially China. China is turning out to be such a vibrant and unique ecosystem and a lot of the ideas like chat POET Technologies, whether just going like crazy and generating a lot of value in China where learning a lot and seeing how to bring those kinds of technologies into other markets like in the US. So those three things end up being very important and for me it’s important to have some balance across those three buckets.

David Kruse: That makes sense, it’s a good way to lay it out. So unfortunately, I think we are out of time. I might try to convince you to come back in a year, because I still have so many more questions. But I really appreciate your time Peter. This is a – you have quite a background and what you are doing now is really interesting. So I know you are a busy guys, so I appreciate your time and thoughts and energy around the show.

Peter Lee: Well, I had fun chatting. So thanks for having me on your podcast. It’s been fun.

David Kruse: All right, great. And thanks everyone for listing to another episode of Flyover Labs. I appreciate it and we’ll see you next time.