Debbie Leung

Debbie Leung

University Research Chair, Institute for Quantum Computing, University of Waterloo

By David Zierler, Director of the Caltech Heritage Project

January 14, 2022

DAVID ZIERLER: This is David Zierler, Director of the Caltech Heritage Project. It's Monday, January 17th, 2022. I'm delighted to be here with Professor Debbie Leung. Debbie, it's great to see you. Thank you so much for joining me today.

DEBBIE LEUNG: I'm so happy to be here. Thanks for the invitation.

ZIERLER: Debbie, to start, would you please tell me your title and institutional affiliation?

LEUNG: I'm at the University of Waterloo. I'm affiliated with the Institute for Quantum Computing, and also with the Department of Combinatorics and Optimization.

ZIERLER: Now, the—

LEUNG: My current position is called a University Research Chair. I wonder if that's what you mean by "title".

ZIERLER: University Research Chair, is that an honorific? What does that mean?

LEUNG: The professorships here come with different amounts of teaching and research. The chair position has a higher research component than the norm.

ZIERLER: Now, the Department of Combinatorics and Optimization, is that separate from the Department of Math or is that the Department of Math?

LEUNG: That is separate from the Department of Pure Math. We have a faculty of math in the University of Waterloo consisting of five departments.

ZIERLER: Now, the fact that it's separate from pure math, does that suggest that there's more of an interest in applied math in your department?

LEUNG: Combinatorics and Optimization are certain forms of applied math. But our faculty of math have a separate Department of Applied Math.

ZIERLER: How does all of that work together? What does that mean to have a separate Department of Combinatorics and Optimization, which must be unique? I'm not sure if I've heard of another department that has this title.

LEUNG: The departmental title is unique to the best of my knowledge. Georgia Tech has a program on similar subjects. My department consists of people who embrace some extreme form of interdisciplinary research. We have graph theorists, cryptographers, and people who specialize in optimization problems, and my department made one of the earliest hires in quantum computing in the field back in 2000. So, it's kind of a fun situation.

ZIERLER: Now, in terms of your responsibilities, your research, your writing, your teaching, are you basically split half and half between the department and the Institute for Quantum Computing?

LEUNG: Very much so. Teaching goes with the department, and research is mostly done within the Institute for Quantum Computing.

ZIERLER: How big is the Institute for Quantum Computing, and when did it get started?

LEUNG: It started around year 2002, and it currently has about 30–35 faculty members (there are people on leave so the number is a little confusing), with 7 in mathematics and computer science (used to have 10 to 11), then a large number of physicists working on both the theory and experimental sides, and then some engineers and some chemists.

ZIERLER: Debbie, I'm curious how this works in Canada. Of course, in the United States, I'm familiar about funding sources. Who supports the Institute for Quantum Computing. What are the agencies or even companies that support the Institute?

LEUNG: There was an initial endowment by Mike Lazaridis, the founder of Research in Motion that created the BlackBerry, the oldest smartphone. The institute has also been heavily funded by the federal government, the provincial government, the university, and grants obtained by the faculty members. So, it's a large number of resources combined so that a large number of people can work within the same building towards the same goal. This started 20-some years ago, a little ahead of many other places.

ZIERLER: Debbie, just as a snapshot in time, what are you currently working on? What are some of the research projects that are interesting to you right now?

LEUNG: Certainly, I've been focusing on one problem that is the channel capacity problem.

ZIERLER: What is the channel capacity problem?

LEUNG: Say, you have a channel that communicates quantum data from a sender to a receiver, and the channel can be used many, many times. It is memoryless from one use to the next (in a simple example the sender sends many bits to the receiver, and each bit can be transmitted incorrectly with probability 1%, independent of whether the bits sent before are erroneous or not). The goal is to find good error-correcting codes with a high communication rate from the sender to the receiver while maintaining very high accuracy of the transmission, and the best rate is called the capacity. The capacity problem is largely unresolved for quantum channels because there are surprisingly interesting quantum error-correcting codes that outperform our expectation, which is exciting, but then it is very hard to show a rate is optimal.

ZIERLER: Debbie, what is your sense of where quantum computing is right now? First of all, do we have a real quantum computer yet, and, if not, what are we waiting for? What are you looking for to make that transition?

LEUNG: It depends on what you mean by a real quantum computer—a small one, certainly. How big a quantum computer do we have right now? There is no agreement on what's the record. It depends on how functional you want every piece to be. [laugh] For a fully functional quantum computer, the largest built may have around 20 qubits or so and "fully functional" may be an overstatement. If you want more qubits—if you can tolerate limitations or restrictions to the operations, or higher noise rates, it is reasonable to say that we have a real quantum computer of about 100 qubits right now. These are big engineering achievements, but you cannot do very much on these small quantum computers in terms of solving hard problems.

ZIERLER: What are the prospects for moving beyond that number? Where do we need to go before we get true quantum computing?

LEUNG: Again, the solution is also error-correcting codes [laugh], error-correcting codes that are resistant to spreading error while you perform logical operation and error correction. This subject—fault tolerance in error correction—is what I used to work on in graduate school, and also on and off throughout my career. These are theoretical ideas to scale up, but we need to be able to implement these ideas in the physical devices. The physical devices have to be big enough because to implement these error-correcting codes, we need enough physical qubits to do error correction. This is a kind of chicken-and-egg problem. [laugh] You need more to provide the protection, and you need the protection to scale up. So, we have to see which side can push through first. [laugh]

ZIERLER: It's interesting that you call this a chicken-and-the-egg problem. To what extent is the problem one a limitation of the theory, and where is the problem a limitation of the experimentation or even engineering?

LEUNG: The experimental side, if I understand it correctly. More accurately, the problems are that of engineering. I once heard that every problem in the lab is just a technical problem. [laugh] I don't want to sound extreme in saying this, but a large segment in experimental physics, and certainly in my field, consists of experiments for which we have a good idea what the outcomes should be, which is unfortunate because there's not very much science left to be found. We basically look for what we think should happen.

ZIERLER: Debbie, to what extent are there real problems out there in the world for which quantum computing could offer solutions, and how might imagining those solutions be useful in getting beyond these technical challenges?

LEUNG: These are two separate questions. We want to find good problems to solve with a quantum computer. That is quite a challenge, despite all the hype on how much quantum computation can achieve. So far, we do not have a large number of problems that we think can be solved much better using a quantum computer. There are only several that are natural problems (for example, factoring, some limited problems in linear algebra, some recent work in machine learning based on shadow tomography, and simulation of quantum systems), and other problems have harder-to-quantify quantum speedups. To build a quantum computer to solve these problems is a separate problem, which is also really difficult. [laugh] Incidentally, I work on neither of these problems. [laugh] I work on error correction which may be indirectly connected to these problems. Most of the time, I am only studying the nature of quantum information. [laugh]

Now, I want to borrow someone else' solution to answer one of your questions in more detail. On quantum algorithms, if you google Steve Jordan (who is also an IQIM postdoc alumni) and "algorithm zoo" you will find a website that lists all the quantum algorithms known, what problems are solve and what are their complexities. It provides a very organized update on the state of art. On your other question concerning how to get beyond these experimental challenges to build a quantum computer—it is always complicated. The experiments are very complex even for a simple calculation. To make the experiments possible, there are necessary simplifications that may affect subtly the interpretation of what are the outcomes and what can be concluded. It will be an important task for the scientific community to understand what experimental shortcuts are legitimate. For example, I still do not know for sure if fault tolerance has been convincingly demonstrated experimentally. Meanwhile, researchers who are developing algorithms have also been working hard to reduce the computational requirements and improving the noise resilience at the software level. Hopefully the two sides can be bridged soon enough.

ZIERLER: Debbie, as you well know, there are major corporations that are supporting research into quantum computing—Microsoft, Apple, IBM, Google, the list goes on. Is your research, or the things that you're interested in, particularly aligned with the different ways that each of these corporations are going about attempting to create a quantum computer?

LEUNG: Sometimes yes; sometimes no. I worked at IBM for a couple of years. That's well before the race to build a quantum computer of any substantial size, so there was less pressure back then. From colleagues who have worked in industry more recently, I have some understanding of the culture and atmosphere these days. Each company has strong interest in getting its approach and implementation to succeed, but less interest in other companies' approaches. [laugh] Understandably, because each company invest so much resources to its chosen approach. Since I work in an academic institute, and do not have ties to any experimental group, and my research mostly tries to understand quantum information, I have the luxury to ask a question for the sake of asking the question, and get the answer to satisfy my own curiosity, without pressure on what should be a presumed answer. For example, I can comfortably wonder which of trapped ions or superconducting qubits will become scalable first, while the same question can be more stressful for a colleague working with 100 people on superconducting qubits. As another example, I do not share the excitement over the immediate-term effort to demonstrate quantum advantage on artificial problems. But these companies also heavily support excellent fundamental research and in this regard, there is great alignment with my own research, or my own wish for where the field is heading.

ZIERLER: To the extent that these corporations are in a race, the race is also to, some degree, international. There's important quantum research going on all over the world. Do you feel like you are part of a larger Canadian effort to achieve quantum computing?

LEUNG: Certainly, but indirectly. First, I'm involved in teaching and mentoring of the Institute's large body of graduate students who have become a substantial part of the global workforce in both academia and industry. Many are Canadians, but only some stay in Canada. But I like the idea that we [the Institute] contribute good researchers to the global community. Within IQC, several colleagues are building small scale quantum computing devices with trapped ions and superconducting qubits. The faculty members train students as a team, and we also have regular scientific exchanges and hopefully these will turn to collaborations.

On the Canadian effort to achieve quantum computing ... twenty years ago, Canada had a much bigger share of researchers in the community. The researchers were mostly distributed between IQC with 10+ faculty members, and several in each of University of Montreal, McGill, University of Calgary, and University of Toronto. Currently, the 30+ faculty members at IQC do not make a big group, compared to the 100-150 permanent staff member at IBM with unlimited [laugh], comparatively, close to unlimited funding and resources. They are not hiring grad students who are still learning the trade, but rather hiring some of the best graduates already trained. There is less teaching or service burden in industry. Building a quantum computer is resource-intensive in terms of researchers' work hours, so competing with, say, Google or IBM, is tough on the Canadian academic community. So, we do what we can do. We do what we're good at.

There are other research directions we contribute more to. One is neutral analysis on the state of field, or scientific methods for doing so. For example, the Institute develops bench-marking techniques for building blocks for quantum computers. These research directions are less resource intensive. We also turn to smaller-scale applications, say, quantum sensing, or very fast switches in the superconducting setup, rather than building an entire quantum computer of a certain size. Research on the theory side does not suffer as badly. So, there are specializations that we can continue to keep our standing. [laugh] But it is a little—it is really tough.

I can elaborate on the scale of the problem. We're really well-funded here—for the Canadian standard. But my funding level typically support one grad student, which is common for mathematicians. My experimental colleagues are better funded. But, still, we have to be strategic in this new age of global competition against setups like those at IBM or Google. [laugh]

ZIERLER: Debbie, to the extent that the creation of quantum computing is going to need advances from all different kinds of researchers, how do you see your interests, your area of expertise, what your group is doing as contributing to that broader effort?

LEUNG: We have to ask what we are computing. This quantum information, which is around for as long as quantum mechanics has been around for 90-plus years, is not much understood. [laugh] I think a basic understanding of what quantum information is hopefully something that will benefit everyone.

ZIERLER: Well, Debbie, let's go back now, and trace your education. This is special because you spent time at Caltech at two different points in your career. Let's start first as an undergraduate. Where were you before you got to Caltech as a freshman?

LEUNG: I grew up in Hong Kong. I was in two separate Catholic schools for 13 years, and then I moved to Caltech as an undergrad.

ZIERLER: Tell me about your desire to come to Caltech to study. Did you know about Caltech from Hong Kong?

LEUNG: Yes, from my high school physics teacher.

ZIERLER: Was the plan for you to focus on physics? Is that what you wanted to pursue at Caltech?

LEUNG: Yes, so that I can understand cosmology.

ZIERLER: Cosmology?

LEUNG: Yes. [laugh] The idea was to get to a place where I could learn astronomy and physics, and then, eventually, I could understand cosmology.

ZIERLER: Now, did you interact with John Preskill at all as an undergraduate?

LEUNG: Very much so. This is the amazing part of being at Caltech—students are welcomed to talk to the faculty members. He was my professor for physics 106 (classical mechanics). I think I did badly. [laugh]

ZIERLER: [laugh]

LEUNG: He's also the PhD advisor of two of my good friends, who were also students from Hong Kong. I knew who's their advisor, who happened to be my professor as well. [laugh] As the professor for the course, John was very fascinating. It's a very engaging course. I was hoping to do my senior thesis with him, but it was before or during his switch from high energy physics to quantum computation research—so, at the time, he told me that I should work on a subject that has problems suitable for undergraduate research students. So, I did my final year project with Professor Steve Koonin, which worked out very well. But John was always a very supportive figure in Lauritsen. I hung out and studied in that building a lot, on the fourth floor library that was in the open but that was very quiet because no one ever seemed to use those journals [laugh].

I got advice from John on where to go for grad school, and he was very generous. If a student walked in his office, and asked him a question, he would help. His other big influence on me was introducing quantum computation to those around him. He and one of his former students, who is a good friend of mine, Hoi-Kwong Lo, were both interested in quantum information at the time. I believe that they influenced one another. So, I learned a little bit of the subject from casual conversations with Hoi-Kwong. John had a website already, with links to interesting papers on the old preprint arXiv xxx.lanl.gov. So, as a final-year undergrad and first-year grad student, I was reading the arXiv papers because of the way John shared this subject to the world.

John also invited Seth Lloyd to give a physics colloquium around 1995. Even now, I remember the talk vividly: this professor from MIT jumping up and down [laugh], writing on the board, talking about some very strange but curiously interesting things about spins. [laugh] So, that impression of the subject stayed with me when I went to Stanford. I was hoping to do cosmology there, but I continued to read about quantum information just for fun. [laugh]

ZIERLER: When you were an undergraduate, John was very much already starting to think about quantum information'

LEUNG: Yes, and he shared on his personal homepage interesting arXiv papers that I read, during the last few months of my undergrad years and the first few months of my graduate studies.

ZIERLER: So, you went to Stanford not intending to pursue quantum information? That was not the original plan?

LEUNG: I went to Stanford intending to work on cosmology, or the blackhole information paradox, which I heard from a physics colloquium given by John. If I were to pursue quantum computation, I would have stayed at Caltech. [laugh]

ZIERLER: [laugh] At what point did you change course at Stanford?

LEUNG: Around April in my first year, so about seven months after starting graduate school.

ZIERLER: Did you think about returning to Caltech? Were there people at Stanford that you thought you could work with?

LEUNG: Yes, I did consider returning to Caltech. My primary mentor Isaac Chuang was initially also a student at Stanford [laugh], and he was about to graduate when I started early 1996. We both worked with our advisor Professor Yoshihisa Yamamoto, but Stanford was isolated from the rapidly growing quantum research community. In the summer, Isaac took me to a major workshop in ITP (now KITP) UC Santa Barbara. We ran into John [laugh] and met his entire group, and we stopped by to visit Caltech. Professor Steve Frautschi realized that I would be quite isolated, and kindly suggested that I could visit Caltech and to learn from John's group—and I did. For a few weeks, I could talk to many others, including John's visitor at the time Alexei Kitaev, John's student Daniel Gottesman, whose work has been a foundation of mine, and John's postdoc Chris Fuchs, whose love of quantum information is forever infectious. It was a very open community; everyone was kind and welcoming, and I was comfortable talking to and learning from them. The fact that the invitation came from John, that he supported a beginning student [laugh] to spend time with his group for her benefit, was a very generous act.

ZIERLER: Debbie, I wonder if thinking back, since you were really there at the very beginning of the development of the field, what was initially exciting about quantum information, both theoretically and looking out into the future, possible real-world applications for quantum information?

LEUNG: For some, the initial excitement was a fundamentally new, theoretical, way to understand quantum mechanics, and for some the excitement was the implications for information processing. For example, quantum data cannot be cloned, and the implication is that we can make secure quantum money (result by Wiesner) and distribute secret key (result by Bennett and Brassard and Ekert). As another example, the dual results of teleportation and superdense coding tell us how to inter-convert quantum and classical data using entanglement and clarifies how quantum data can be split into parts, paving a method for quantum error correction, despite the no-cloning result. These results were complementary to the equally exciting invention of quantum algorithms and the quantum circuits model, affirming that quantum computation is digital and quantum noise can be digitized, so that quantum noise would not be a fundamental obstacle in realizing quantum computation.

And the more we understood quantum information, the more surprised we were and the more questions we had, and attempts for further clarifications gave rise to more interesting phenomena and applications.

As a beginning grad student at the time, I just felt that even the most basic foundation of physics and quantum information was not understood, so, why not spend a few years [laugh] or a few decades [laugh] getting that sorted out?

ZIERLER: Debbie, was there a sense that advances in quantum information would be useful for physics itself?

LEUNG: Very much so; tools and ideas from quantum information have been used in some accelerated manner in theoretical physics in the past two decades. One example is the result by Patrick Hayden and John Preskill on the black hole information paradox problem. (Patrick was my officemate when we were both postdocs at Caltech.) They applied a quantum information argument to derive a complexity viewpoint on why the paradox may not be as big a paradox as we thought because one cannot decode fast enough to create two copies of the same quantum state. Another example is the work by Harlow, Pastawski, Preskill, and Yoshida, modelling the AdS/CFT correspondence in high-energy physics using quantum error correcting codes. (This was another Caltech production with Pastawski and Yoshida being postdocs and Harlow a visitor there at the time.) Another faculty member at Caltech, Fernando Brandão, whose thesis was on entanglement measures, made important applications to condensed matter physics, in quantifying correlations measured in physics experiments. I think there are many information theoretical questions that maybe physics cannot run away from. [laugh] Also, other communities may have been too busy to sit back and fully resolve some questions that arise in their work, but people in quantum information often study quantum information for the sake of it [laugh], and we are happy that our answers are useful elsewhere. This is how I view the relation between quantum information and other branches of theoretical physics.

ZIERLER: Debbie, given the situation at Stanford, what was the process like for you developing your thesis research, your dissertation?

LEUNG: It was pretty difficult [laugh] because during the entirety of my thesis work, for four and a half years, many of the faculty members at Stanford did not believe that we could build a quantum computer. Their instincts were based on other models of computation with unfulfilled promises and the concern of quantum decoherence. We spent a lot of time explaining the most basic ... very basic ideas, such as why quantum computing is not analogue, why quantum errors can also be discretized, and how quantum error correction can be done, but it was difficult to fight prejudice. [laugh] We had to convince the rest of the campus [laugh] we're not lying, and our work was solidly thought out. Having to find readers for my thesis was another headache. [laugh] It wasn't easy, I'll put it that way. It's quite amusing that, in the past few years, we have to do the opposite—to calm people down, [laugh] and remind them to not get over-excited by the promise of quantum computation. [laugh]

ZIERLER: Right, right.

LEUNG: Back in grad school, it took serious effort to reassure others that we knew what we're doing—not that everything was already working in quantum computation, and there were a lot of difficulties ahead, but the most important one, two, three foundational ideas were carefully worked out.

ZIERLER: Now, was Yoshihisa Yamamoto, was he working on quantum information at all?

LEUNG: He was. He was one of the earliest. His research on experimentally observing the ultimate quantum effects brought him to quantum noise in computation and communication as early as 1986-1988, and he and Isaac Chuang were proposing qubit systems and quantum error correcting codes when I joined the group. He was deeply involved in most of my work. He's very supportive. But it's just the three of us against the rest of the campus for years. [laugh]

ZIERLER: Now, in the late 1990s, early 2000s, of course, this was the height of the dot-com boom and so much growth in Silicon Valley. Was anybody in Palo Alto, in Silicon Valley, was anybody interested in what was happening in quantum information at Stanford or was that too early?

LEUNG: That was too early. Even within the academia, quantum information processing was considered risky research. Isaac Chuang was hired as a permanent staff member at IBM Almaden Research Center, and other students at Stanford and I would visit him there. That was an extension of the interest at Stanford and IBM. Umesh Vazirani and Birgitta Whaley at UC Berkeley were neighboring groups we would visit. The first startup company in quantum information, to the best of my knowledge, is MagiQ in Manhattan. That was built around 1999. I still remember teasing Hoi-Kwong Lo (who was the founding director for the company) in a workshop in Cambridge University that someone was actually privately funding [laugh] quantum research, only to find that the person who funded that research was the stranger at the same table. [laugh] The company was developing products in quantum key distribution and security, and so it was a bit embarrassing on my side in hindsight. [laugh] A little later, there was a second company ID Quantique. Both were fairly small companies on quantum cryptography, an area requiring very little quantum manipulation.

ZIERLER: Now, robust, what did you mean by the idea of robust quantum computation, as your thesis title suggests? What were the principal conclusions or main arguments of your thesis?

LEUNG: At the time, the community had developed some methods for error correction and computation with imperfect components. But these theoretical methods require a lot of resources which we still don't have to-date. My thesis provides a collection of ideas to suppress noise with very, very little extra resources, as first steps towards error correction. For example, our error-detecting code only had two spins in it [laugh], so we did what we could.

ZIERLER: Now, the idea of thinking about robust quantum computing 20, 25 years ago, what are the gains? What is now robust that was only dreamed of being robust when you were a graduate student?

LEUNG: Oh, my! [laugh] Wow. [pause] Quantum gates can be 99% [laugh] accurate nowadays, which was unimaginable back then when 80% was pretty amazing. The "quantum computers" are also much bigger now. When Ike (Chuang) told me in '97 that "I have a two-qubit quantum computer. What experiment are you proposing?" [laugh] I said, "Can you run error correction? I want to see error correction." So, we crammed this teeny little error-detecting code into the 2-spin (two qubits) molecule because we did not have the three qubits needed for error correction. We had to do quite a bit of work to prove that we have improved the error. Last year, Chris Monroe announced he had 22 qubits in his ion trap. [laugh] I am still waiting to see various notions of fault tolerance demonstrated, but it's really different now what people have.

ZIERLER: Now, tell me about your initial postdoc at IBM. Was IBM thinking about quantum information at this point?

LEUNG: Yes! Researchers at IBM like Rolf Landauer and Charles Bennett in the Physics of Information group were asking related questions much earlier (in the 60s and 70s): Can you perform computation reversibly? Can you perform computation without using energy? Can you do this or that? So, the question "if you have quantum information, what can you do with it?" was part of a bigger theme in the group. As early as 1982, Bennett, Brassard, Breidbart, and Wiesner proposed a quantum method to reuse a secret key, which further inspired the first quantum key distribution scheme by Bennett and Brassard in '84. Bennett co-invented superdense coding and teleportation in the early 90s. After Shor reported his algorithm in '94, the group including Charles Bennett, David DiVincenzo, John Smolin, Barbara Terhal, and their collaborators developed a large number of results on how to perform quantum gates, error correction, entanglement purification, and discovered other exotic quantum effects, and so on. They were there throughout. [laugh] The group had a few people, like three, four, or five, for many, many years.

ZIERLER: Debbie, did you feel that when you arrived there that there was maturity in the thinking at IBM that was not present elsewhere?

LEUNG: They were really ahead [laugh], just like at Caltech. There were very few places that were really ahead. The group members knew vastly more about the subject than reported in their numerous papers. One can attribute much of the major development of entanglement theory and quantum Shannon theory to the group. Their playful curiosity was inspirational.

The manager at the time, Nabil Amir, led two groups, one at Almaden near Stanford, and one at T. J. Watson near New York. He gave an opening speech at a workshop in 2000 that he couldn't care whether we built a quantum computer in the end. [laugh] He got great people and they would get nice results, whatever they were. [laugh] This was in the Physics of Information group, [laugh], and studying quantum computation was certainly a great way to learn about the physics of information. Of course, the group was confident that quantum computation would be a promising model and it was not a surprise that they had a strong involvement throughout. But, back then, "strong involvement" meant a five-researcher group including postdocs.

ZIERLER: Debbie, I'm curious, at IBM, was the research environment one where IBM was supporting simply fundamental research? Was there any interest in achieving results that might be at some point valuable, economically valuable to IBM?

LEUNG: As a postdoc, I was never involved in the reporting or how the group was being supported. I only had to do research and write papers. [laugh] So, I have to speculate here. At some level, say, my manager's manager's ... to the fourth power [laugh] might have to explain to some big boss why this research had to be done at IBM [laugh], so some justification on the potential and the promise must be around at a very high level. Lower down, my group and my immediate manager were doing fundamental research without having to claim economic value. My immediate manager and his manager likely talked in terms of the fundamental science. [laugh] So, I think at many levels of the management fundamental research was protected—at least when I was a postdoc there. Also, it might not be easy to distinguish fundamental from applied research. So, there were pockets of time and space that one could just hide in and do what they knew was fundamental research. But, at many levels, one could also explain to others why there could be potential economic benefit. The two narratives are not exclusive.

ZIERLER: Now, what aspects of your work at IBM were a continuation from your thesis research, and what were simply brand-new projects to you?

LEUNG: It was quite a sharp change. [laugh] When I was a PhD student, I asked theoretical quantum information questions that were inspired by, or were attempts to address potential experiments. When I moved to IBM, the experimental group was 28 [laugh] aisles away. [laugh] I largely detached myself from experiments and switched to work on purely abstract quantum information. So, I was working with the group on entanglement and capacity and cryptography and anything I wanted to study like measurement-based quantum computation. [laugh] It was a very sharp change, and I continue to work on these topics to-date.

ZIERLER: Was your time in Berkeley simply a holdover before getting to Caltech? What did you accomplish when you were at Berkeley, since it was so short?

LEUNG: Yeah, that's right, MSRI at Berkeley had a 4-month workshop that was between my two jobs. It was tricky to sign up either employer to be my official employer while I was away. [laugh] So, MSRI offered me an actual postdoc position for four months. It was an exceedingly productive stay. A lot of the research I did in my first year at Caltech started at MSRI. The work on composability of quantum key distribution was started by discussion with Dominic Mayers. The line of research I had with Patrick Hayden and Andreas Winter on randomized arguments for quantum information also took off at MSRI.

ZIERLER: Now, were you in touch with John Preskill while you were at IBM? Were you following all of the exciting developments that he was involved in?

LEUNG: Certainly. The field was so small, and exciting things are exciting things for all! [laugh] We're not going to miss out on someone else's exciting things. [laugh] Conferences and workshops brought most of the community together several times a year. I was also visiting Caltech when I worked at IBM—in May 2001, and that was a little before I applied for a postdoc at Caltech.

ZIERLER: Now, were there other programs that you applied to, or you were really focused on getting back to Caltech?

LEUNG: I applied to many places because it was not easy to get any job in quantum information. We could not afford to apply to only one place. [laugh] We cast a wide net. But Caltech was my top choice, so I was very glad to be back.

ZIERLER: In what ways did it feel different at Caltech with the creation of IQI from your undergraduate days?

LEUNG: We had a hallway with 10 postdocs [laugh], two in an office, five offices in a row, in Steele, and that was nice. [laugh] Two more sat in Jorgensen, so there were 12 of us, and there were a bunch of students. They were very good students! It was a very productive community. There is a joke that I can make even [laugh] on record. John paid for the students. [laugh] He took all the risks [laugh] accepting the students. The students worked with us (the postdocs). These were really bright students. They had very good ideas. They wrote the papers. So I had some of my best experience working with students who formally were John's but they worked heavily with us. Another really nice aspect of the workplace was the nice peers all about the same age and with similar attitudes. Most of us went through PhDs with more isolation. [laugh] There were not many places that heavily supported quantum information research at the time. So, we were in this place where there's a team. [laugh] So, that was very nice. Also, John was there and Alexei (Kitaev) was there. So it's like an amusement park—every day was a joy ride. [laugh]

ZIERLER: [laugh] Debbie, were you the only woman in the group at that point?

LEUNG: Technically, no. For the first two years, I had overlap with—Charlene Ahn, one of John's students. For the last year, I had overlap with Hui Khoon Ng, also John's student. Sorry that I needed to struggle to recall their names. I should be able to remember better [laugh] when there was only one other [laugh] woman in the group at any time. The two students had no overlap, and their offices were far from mine and there happened to be no collaboration. Ann Harvey, our admin staff, and I, bonded well.

ZIERLER: Debbie, with all of this productivity among the postdocs, where was there competition and where was there collaboration among you and your peers?

LEUNG: If I may say that, it's all collaboration. This was the fun part in the field at that time. The job opportunities were so rare that it's pointless to even think about competition. [laugh] It was not productive. We also believed that it might not be a zero-sum game. Patrick and I, we shared an office. We applied to the same places. We shared information on jobs. [laugh] We even thought that we should apply together [laugh] ... that we should both apply, when we saw a position that was not in quantum information, say, in general physics, just to see if the recruiters would change their mind. [laugh] That quantum information was a field that they should consider. [laugh] So, it's all very positive and collaborative, and we got the much needed peer support. Not peer pressure, but certainly the postdocs motivated one another.

ZIERLER: Was anybody thinking about possible connections between quantum information and LIGO, which was starting to get really exciting around this time?

LEUNG: I remember seminars presenting ideas to improve the sensitivity of LIGO using squeezed light or interferometry, with more quantum information theoretic understanding, but I cannot locate any reference.

ZIERLER: Now, you were what was called a Tolman scholar, a Tolman fellow. What does that mean at Caltech?

LEUNG: I believe that means my position was funded by an endowment related to, or directly from the Tolman family. A number of former Caltech postdocs in quantum information had named fellowships under Sherman Fairchild, or Tolman, or DuBridge. These private funders must have wished to support fundamental research and their names are attached to the positions. The other unnamed postdoc positions probably have been funded from grants and other sources.

ZIERLER: What were some of the big ideas that were animating IQI at that point? What were people really excited about?

LEUNG: A LOT of things! [laugh] John was working on the fault-tolerance and threshold theorem using a distance-3 code, with former student Daniel Gottesman and then student Panos Aliferis. Dave Bacon, who is now at Google, was working on a large set of problems like how you can get better error-correcting codes with gauge bits, giving a major method for fault-tolerant gates these days. He was also trying to use closed time-like curves [laugh] to solve NP-complete problems. Patrick and I, sharing an office, and Andreas Winter, who visited us often, were working on entanglement properties of generic states, and using randomized techniques to prove unexpected communication capacity results with others. That's what we did. Robert Raussendorf next door was working on fault tolerance methods in measurement-based quantum computation, and extensions to other topological solutions. Sergei Bravyi and Alexei Kitaev developed the magic state distillation method for fault-tolerant non-Clifford gates. Guifre Vidal switched from entanglement theory to DMRG method for simulations for quantum systems. Many of us still continue on our line of research from those days.

ZIERLER: What was John Preskill's presence like at the IQI? I mean, was he always around? Was it more formal? Did you need to make an appointment to see him? How did that work?

LEUNG: [laugh] John's presence changed over time. I visited many times after I moved to Waterloo, and John was joining the group lunch with students and postdocs almost daily in the later visits. [laugh] While I was a postdoc, we mostly saw John at the weekly IQI seminars and the group meetings, and only occasionally during lunch. We could talk to him directly without an appointment; he was very supportive, though we'd rather not disturb him, assuming he must be very busy.

Most discussions with John happened over group meetings. The group meetings were weekly, lasting 2-3 hours, with dinner and tidying afterwards. We went around summarizing our progress, getting feedback from one another. There was a strong sense of community. I really looked forward to the group meetings. There was something rather funny about John's presence [laugh] ... I have to find the right words. We had some anxiety because he's this figure that we revered tremendously. We worried that—he would be skeptical about our work. [laugh] He's very, very generous and very approachable, so this anxiety was strange, but very real and prevalent. Perhaps due to our respect for him, we wanted to "pass"—if I can use those words. [laugh] Very few of us, I think actually none of us, collaborated with him at the time. Postdocs were working with one another and with students, but not formally with John at the time. So, perhaps our lack of knowing John outside of the group meetings added to this anxiety.

ZIERLER: Now, this idea that this would create anxiety when you're presenting your research, how would John convey whether he thought you were on the right track or not?

LEUNG: I learned that already the first time I gave a talk at Caltech in '97. My mentor Ike Chuang told me that if John stopped taking notes [laugh]—

ZIERLER: [laugh]

LEUNG: —you lost. [laugh] You lost him. [laugh]

ZIERLER: [laugh]

LEUNG: He took detailed notes in all seminars and group meetings. [laugh] So, if he stopped taking notes, oops [laugh]—

ZIERLER: Do you have any recollections of thinking, "Wow, he's still writing. I must be onto something good here"?

LEUNG: [laugh] Well, yes, at times, especially when there were additional questions, feedback, or suggestions for extending the work. We took turns to give the long talks by alphabetical order in our names, and I was supposed to speak the week before John. When I traveled, John would switch with me, and I missed almost all of John's talks. [laugh] But he was always there when I had to speak, and, yes, it's very nice to hear his opinions, questioning about—like, whether our focus was a good direction, whether we missed one crucial aspect of the entire research. He would ask you, "Have you thought about this?" "Well, sorry, no." [laugh] "Yes, certainly we should have." [laugh] "It's certainly important." [laugh]

If I may add this, there were really playful moments in the group meetings. When the meeting ran too long, John flipped a Canadian toonie to see if the half room represented by the bear or the queen should report. Once, we all sat on one half of the room to defeat this ritual. [laugh] Also, in an egg-and-chicken manner, John always ended up in one specific big green sofa. When the group moved to Annenberg, the green sofa was lost! Steph Wehner, postdoc at the time, told of a story in John's 60th birthday celebration, how she relocated the sofa in a storage room on campus and organized a "recovery" of the sofa in the new group meeting room. [laugh] I would say John's presence was very appreciated, feared, and loved all at once.

ZIERLER: Debbie, of course, this was long before the creation of IQIM. Were there condensed matter people that were hanging around IQI when you were a postdoc there?

LEUNG: Not that I remember. There might be one or two, but I do not have very clear recollection of that.

ZIERLER: Is your sense that this is simply because there was so much theoretical work that needed to be done before we can start thinking about merging matter into the IQI?

LEUNG: Yes and no. After the big years for quantum computation in '94-'96, it took another 10 years or so for quantum information ideas to be used in other fields. It might be that indeed we had to grow up first. [laugh] When we were grad students, we had to learn quantum information on our own. There were no textbooks, very few courses or peers, so most of the time, we had to learnt from the original publications with more effort. It took more time to stay on top of our research. Later, there were more textbooks, more hired as professors [laugh] who could advise and teach graduate courses in more places. At IQC, we started a graduate program. Students and postdocs spent less time re-deriving existing knowledge, so they had more time and maturity to learn the major problems in other areas. Also, more students working in other disciplines have solid background in quantum information.

Also, other communities started to accept quantum information as part of physics or part of cryptography etc. Before that, many were hostile or dismissive. Someone told me at a cryptography meeting in 2005 "I just hope that no one ever build a quantum computer, so I don't need to worry about quantum anymore." Or in a physics department, people would tell me, "you guys will never build a quantum computer." Some people just wanted—for years—they just wanted us to disappear [laugh]. It took a while for more researchers to accept that quantum information people are legitimate [laugh] scientists ...

ZIERLER: [laugh]

LEUNG: and they may even be helpful.

ZIERLER: [laugh]

LEUNG: So, researchers in other fields were increasingly more willing to attend our talks and talk to us [laugh] about what they're doing. Once they thought that quantum information people might have useful ideas [laugh], they're more willing to share their problems with us, and then we solved some of their problems.

ZIERLER: Debbie, what do you see as some of those problems for which quantum information scientists started to be taking seriously or more seriously by their academic colleagues? What were those problems?

LEUNG: One example is the blackhole information paradox. Here is one insightful thought experiment. Throw a piece of quantum data into a blackhole, which subsequently go through Hawking radiation. How much Hawking radiation has to be omitted so that the infalling quantum information can be recovered outside the blackhole? First, it helps to model the in-falling quantum data as rapidly mixing with the existing blackhole—and this is similar to encoding quantum data into a random error correcting code, a common technique in quantum capacity calculations. Second, recovering the quantum data is similar to the decoding step of capacity studies. Third, initial correlation across the event horizon and success criterion for recovery are well studied in quantum information. Researchers in blackhole information paradox such as Leonard Susskind discussed many of these issues with the quantum information community and Hayden and Preskill answered the question above.

Another example is modeling the AdS/CFT correspondence. Almheiri, Dong, and Harlow initiated connection with the three qutrit erasure code, and when Harlow visited Caltech and presented the results, he and the IQIM postdocs Beni Yoshida and Fernando Pastawski and John Preskill found a family of exactly solvable toy models for the correspondence capturing many more useful features. So, they used error correcting codes to model the spacetime. [laugh] To be fair, in this case, Daniel Harlow is bilingual in high-energy physics and quantum information, but he certainly is a high-energy physicist using quantum information and collaborating with quantum information theorists.

There are other examples in condensed matter physics and quantum chemistry, and Fernando Brandao, Xi Chen, and Garnet Chan at Caltech may know more.

ZIERLER: Debbie, looking back, what do you see as your most significant work as a postdoc at Caltech? Why was it so significant? What was significant about it?

LEUNG: Patrick Hayden and Andreas Winter and I realized that certain randomized techniques were extremely useful for solving many questions in quantum capacities and entanglement theory that we were asking. We adapted and developed those techniques for quantum information, and also solved the problems we were interested in. The approach also attracted a lot of follow-up research and extensions.

We found strong "measure concentration" for many quantities, meaning that as the system size increases, despite the enormous possible complexity of the state space, overwhelming fraction of the states have similar value for these quantities. We used this to calculate and understand generic or average behavior. In addition, we could quantify how overwhelming the average effect is, giving us methods to prove what can and cannot be done in quantum data transmissions, which is why these randomized techniques solve capacity problems.

One result coming out of this is that it's really hard to not have entanglement, opposite to our intuition. Not having entanglement is an accident. It's a very, very unlikely event, and we know how unlikely. [laugh] This result is interesting on its own, but it also gives us a method to transmit an arbitrary pure quantum state that is known to the sender using entanglement to replace half of the communication, improving the method for superdense coding for quantum states Patrick and Aram Harrow and I studied earlier. We found many other interesting results on quantum correlations.

Another result coming out of this body of work (with additional collaborators Charles Bennett and Peter Shor) is how much it takes to randomize an arbitrary quantum state; by this, I mean someone applies one of many possible changes, and can reverse the change and recover the state, but anyone else not knowing the change finds a completely random state. This is a quantum generalization of encrypting data, with the secret (key) being the knowledge on what change has been made to the state. We found that allowing an arbitrarily small approximation reduces the required key to half of what's otherwise needed, if the eavesdropper has no side information. The same mathematics allows a different problem, which is analogous to teleportation, but with the sender knowing the quantum state to be sent, to be solved with half the communication. These schemes are optimal, and the savings are surprising, because it happens very abruptly with just a little approximate. This result also implies the difficulty of process tomography and provides schemes for hiding correlations.

Subsequent development on randomized techniques by others include how Matt Hastings disproved the 4 equivalent additivity conjectures that were opened for 12 years.

Research is always evolving and how we think about them is changing too. But the point is that we keep learning. A personally significant outcome of my Caltech postdoc was a switch to more heavy-lifting mathematics with peer support from my collaborators. I have an undergrad degree in math [laugh] and in physics, but many aspects of my mathematical background turned quite rusty by the end of my first postdoc. Re-entering this world was challenging but fun. It also enabled me to switch to the more mathematical side of quantum information.

ZIERLER: Debbie, by 2005 when it was time to start thinking about your next position, just as a personal question, at this point in your career, was returning to Hong Kong, was that not something you were interested in?

LEUNG: That's a tough situation. I wanted to go to a place where there would be other colleagues in the field. Having more people in the same field would mean a lot.

ZIERLER: Now, either in Hong Kong or Mainland China, were there such places that you considered that might have been feasible for you?

LEUNG: Not particularly. I was hoping, I mean, it has always been on my mind to return to Hong Kong. I'm a proud Hongkonger. [laugh] But research is also very important. I have multiple identities. This quantum researcher identity is so important to me that I want to be associated with it my whole life. When I was looking for a permanent job, I was really hoping to pursue quantum research as much as I could, so that's why I picked Waterloo.

ZIERLER: What opportunities were available to you? What was interesting as you looked beyond Caltech?

LEUNG: You meant around 2005—

ZIERLER: Right.

LEUNG: I would be very content with a permanent job where there were colleagues with similar research interest. We're lucky that by the end of 2003, a few places started recruiting in quantum information. Before that, candidates in quantum information basically needed to plow their way to land a job. It's like, basically, when you need to climb Everest, you have to break your own trail. They applied to jobs that were general searches or those targeting areas outside of quantum information, either way they were uphill battles. The recruiting departments might not have anyone knowing quantum information research, and there were many skeptics of our field. By the end of 2003—the trail [laugh] was still rough, but there was a trail! We applied to places where they specifically look for quantum people, so that's a big change. [laugh] I interviewed in 4-5 places, most were specifically hiring in quantum. The fact that there were searches tailored at quantum and getting interviews meant quite something to me already.

ZIERLER: What did you end up doing? What was the next option?

LEUNG: You meant in terms of how I ended up in Waterloo?

ZIERLER: Yes.

LEUNG: The decision wasn't too difficult. The Institute for Quantum Computing was being established in Waterloo, and they recruited several people I could work with. These have been close colleagues and collaborator since. [laugh] It's a cool place. [laugh] I spent 12 years in California, with many undergrad and grad school friends there. It would be nice to stay in California and I had an opportunity but with a lesser research environment compared to Waterloo. Research came first, so that's it. [laugh]

ZIERLER: What were the most important projects to you at this next stage in your career? What did you want to focus on next?

LEUNG: Many of the communication problems I considered earlier were concerned with point-to-point communication, with one sender and one receiver, and one large message from the sender to the receiver. After moving to Waterloo, I worked for a while on more general communication settings. These included bidirectional channels modeling two-body interactions, network coding, other form of interactions like free side classical channels, transformations enabled by free side classical channels, or transformations enabled by no communication at all. We found many new quantum phenomena in the process. Then, I switched back to point-to-point communication again, as nudged by new ideas and techniques and opportunities to tackle open problems.

Looking a little into the future, I hope to understand the fundamental problem—for what channel the capacity is positive. It sounds like such a small problem, but we have been stuck for 26 years. [laugh] For example, the channel that takes a bit as input and flip it with some probability p; there's a quantum generalization called the qubit depolarizing channel. You either send the qubit noiselessly or with some probability, you replace it by some "garbage." The tricky part is that the users don't know whether the output is garbage or not. For a range of p about 10-20%, we do not know if this channel can send any quantum data at all, even given unlimited number of uses. We cannot prove that no quantum data can be sent, but we cannot find a method to send even a qubit. It's a huge gap in our understanding. That is outrageous. [laugh] It is rather humiliating. [laugh] I really hope we will resolve this problem at some point.

Last time I talked to John, he was a little skeptical [laugh] that I am fixated on this problem if the difference is zero capacity or almost zero capacity. My justification is that, we must be missing a lot of understanding to not be able to ask such a fundamental question. It's not the answer, per se, that I want to understand. But our lack of method to answer some modest simple-sounding questions is concerning. If we don't understand why we cannot answer these simple questions, I worry that there's some big block in our understanding. Like, we must be missing a chunk of what we think about quantum information, leading to our not being able to answer some simple questions.

ZIERLER: Do you see that, Debbie, in similar terms to the block of understanding how to merge quantum mechanics and general relativity, or trying to figure out how we get beyond the Standard Model? In other words, is this block, as you're describing it, one of many blocks that we experience in physics, or do you see it as unique, separate from those other problems?

LEUNG: There's no obvious connection that I'm aware of. But our understanding is so sparse [laugh] that we don't even know [laugh] whether different problems could be related. [laugh] The block that stops us from understanding capacity could be some aspect of quantum information that actually has something to do with our inability to merge it with gravity.

ZIERLER: Is quantum information waiting for—what?—an Einstein, a von Neumann? Are we waiting for a genius to come around with a eureka moment? Is that what it is?

LEUNG: That's another strange question. [pause] I guess I don't think about such moments in terms of a person, but in terms of some group effort to tackle a notable problem. Say, there is a question that people really want to understand, and the more people try, the harder the problem becomes. [laugh] Then, at some point, suddenly, there will be several very significant results in some short amount of time, leading to a partial or full solution for the original problem and also solving many other problems in the process. After that event, we'll all grow up a little. One recent example in a different topic in quantum information is MIP* equals to RE. Thomas Vidick wrote a beautiful blog on the history. One starting point of this problem sounds specific about non-local games. Is it possible that we need infinite entanglement (in terms of some C* algebra) to play a finite game optimally? This question sounds specific to a small group of people with a little obsession. [laugh] The answer turns out connected to a long-standing open problem in a different field (operator algebra) called the Connes embedding conjecture, and to the expressive power of the complexity class "multi-prover interactive proof system assisted by unlimited entanglement."

Slowly, people uncovered these connections, and pieces of the puzzle were solved. It still wasn't clear how to solve the problem for a while and then, suddenly, a large number of interesting results came together, solving five open problems in one paper, five ridiculously large open problems in one single paper [laugh] early 2020. [laugh] There are five authors and many collaborators in the earlier pieces. Then the community has been learning these new ideas in the last two years, and then we all grow up a little. [laugh] Such breakthrough happens once in a while. When working on such a problem, it can feel uncertain and lonely, with very few colleagues to share and collaborate. It takes a lot of time and effort and persistence. But with these persistent, determined [laugh], little crowds of people [laugh] knowing who else to bring in, knowing roughly a direction, eventually, will again, certainly, bring in big progress to the community. So, pieces at a time, hopefully we will grow for many, many decades to come.

ZIERLER: Debbie, when you arrived at Waterloo, and it was time to build up your group, and contribute to the growth of the Institute for Quantum Computing, what did you learn during your Caltech years as a postdoc, both in terms of collaborating, in terms of thinking about problems? What did you take away from Caltech that was useful to you as you started this work at Waterloo?

LEUNG: I enjoy working with students and peers. I love direct involvement, so I made a decision to keep my group very small [laugh] but working actively rather than overseeing the work from a distance. Another thing I took from my Caltech days is to continue to ask the questions that are interesting. The pressure for grants is a little less severe in my case, especially with a small group, so I feel reassured that I don't need to chase the hype—I wish I could find better words. I work on two to three interesting things most of the time and hope to obtain some meaningful result in at least one of the projects every two or three years, and hope to avoid nonscientific reasons in deciding what to work on.

So, maybe it was a kind of determination and [laugh] reassurance [laugh] that I probably just took from my postdoc days. I'm not sure how that arose. A postdoc should be concerned about jobs and opportunities. I may not represent my former fellow postdocs, but I don't remember anyone being particularly concerned. Perhaps jobs were so scarce that we had low expectation. [laugh] When I entered the field as a grad student, people told me the field might not be there when it's time to write up the thesis. Of course, I knew that the field had good foundations, so that should not be a problem, but there was little expectation to get a permanent job or grants out of good research. So, I just did the research I wanted to do. [laugh] I felt that was a prevalent attitude among my peers too.

I still feel the same way. For example, it doesn't work motivation-wise to guess what projects will get me funded. Time and energy and effort are valuable and limited so it is important to prioritize what we find interesting. I'm preaching here. [laugh] When I came back to Caltech to visit, I felt more encouraged to put aside unnecessary worries. [laugh] Sitting at Annenberg or outside the Red Door café, I just asked "What do I actually really want to do?" In my Caltech days I was protected to pursue things that were important to me, and hopefully important to the field. This is an encouragement that I always take away from my Caltech days.

ZIERLER: Debbie, on that point, because you follow your interests, and because your research is somewhat aloof from the nuts and bolts of creating a quantum computer, how do you measure progress? How do you know that day in and day out, you're making progress, you're doing things better, you're going towards something that is bigger than where you were before?

LEUNG: Instead of counting qubits in the experiments, using less time or space in an algorithm, improving error thresholds, progress in my field can be measured in other ways, for example, by how much we have extended the parameter range or the settings for which I can answer my questions, or tightened bounds on the quantities we cannot evaluate. We are making progress on all three fronts for the capacity problem. For an open problem, even without solving it, partial progress can be one step towards the eventual solution, or a much simpler or insightful alternative second or third solution. I can also measure that progress in terms of better understanding of a problem. If we find results that others are interested in, or people use our new ideas or tools to solve their problems, that's a good sign of success as well.

ZIERLER: Debbie, we talked at the beginning about your current work. So, for the last part of our talk, I'd like to ask a few questions looking to the future. Extrapolating what you've done up to this point, how will you define the projects that will stay with you, that you have worked on for a long time, that continue to require you to work on for a long time, and what are the projects that you look forward to pursuing at some point in the future, which might not be possible from a theoretical standpoint yet?

LEUNG: You're asking how do I distinguish these two possibilities?

ZIERLER: Yes, that's right.

LEUNG: [pause] I do not think much in those terms. That's my honest answer. [laugh] I don't want to use this analogy, but, it's like a relation between me and my research. [laugh] I do my research in a way that I can't stop thinking about the problem whenever I'm awake. Therefore, I end up working on it. [laugh] Why for the last 17 years I'm still obsessed or attached to the same problem, I do not have a clear answer.

ZIERLER: It's curiosity. It must be curiosity.

LEUNG: Certainly! But I can have multiple curiosities too. [laugh] The curiosity and attachment shift between several questions at any one point, and which several questions change over the long term. I know why each of these questions are interesting, but there are many other interesting questions I do not work on. For me—there's no way to do research except for [laugh] being driven by this strong attachment. [laugh] It's hard to rationalize attachments. [laugh] Yes, I do not have a good answer to your question.

ZIERLER: How do you prevent yourself from being driven crazy by these problems, working on them for so long?

LEUNG: [laugh] [pause] Maybe our understanding is so incomplete that, so far, there's always a new aspect of the problem to investigate. Every so many months, there will be another new insight or realization that fuels some high hope for at least a few days. [laugh] Maybe we have a chance to grow that understanding again—not all the time but sometimes. I grow to accept that things can be stuck for a while, and if I'm stuck on one question, I switch to the other two or three questions I have in mind for a break. [laugh] I think once in a while then there's an insight, and so there is a way to remain optimistic. Working with collaborators also helps. For example, for the last two years, I have been working largely with Graeme Smith, a former student of John whom I have been working with since 2004 when we're both at Caltech, John Smolin, former colleague from IBM, Felix Leditzky, a former postdoc of Graeme and I, and Vikesh Siddhu, the youngest addition to the cult. [laugh] So, there is a support group [laugh]. We sometimes work like a basketball team brainstorming together, but at other times, like running a relay. We are less alone. For many of us, our collaborators are also our closest friends. We do not always make progress in our weekly meetings, but at least we have fun. [laugh]

ZIERLER: [laugh]

LEUNG: [laugh] Work wise, we are all very different, so we can add to one another's strength to continue to push the project forward. Just like climbing a difficult route, going in a team helps.

ZIERLER: Finally, Debbie, last question, when these breakthroughs are achieved, whatever that looks like and whenever that happens, how will you understand Caltech's contribution at the institutional level, what John has built, all of the postdocs and professors who have contributed to IQI and beyond, what is the sum of what all of that will mean when we have these breakthroughs?

LEUNG: It really will mean a lot. For more than 25 years, Caltech as an institute and John's leadership in QUIC—I think that's the name of the project in the mid 90s, IQI and then IQIM, have provided a wonderful academic environment for students and postdocs to develop their research interest and strength in quantum information, and for the several faculty members at IQIM to thrive. There are 100+ postdoc alumni and a comparable number of PhD graduates, plus additional undergraduates and interns, many of them are leaders of major institutes, or leaders of their respective areas of focus. (I want to give some examples, like Scott Aaronson, Dave Bacon, Andrew Childs, Patrick Hayden, Robert Huang, Barbara Terhal, Steph Wehner, but for every name I put down, I see 5-10 others I should also include; my apologies with the omissions.) Also I think my experience must be very generic. Then we take the learning, the expertise, the collaborations, and the motivation with us elsewhere. That's a large set of [laugh] influence and drive for the field. I will tell you another science story if you have time.

ZIERLER: Please.

LEUNG: Aram Harrow, a close friend and collaborator and now faculty member at MIT, had a funny story how he missed a flight when he started in the field as an undergrad in 2001. He went to the wrong gate, sitting opposite to where he was supposed to board. He was [laugh] reading John's lecture notes, and he was so fascinated by the reading that he didn't hear those extremely loud calls for him to board. He kept reading, and then he missed his flight. [laugh]

ZIERLER: [laugh]

LEUNG: It's hard to measure influence, you know. [laugh]

ZIERLER: That's great. [laugh] Well, Debbie, I'd like to thank you so much for spending this time with me. It's so wonderful to hear your perspective as we're thinking about IQI and IQIM, so I'd like to thank you so much.

LEUNG: Thanks for doing this project too.