Assistant Professor of Computer Science, University of California, Davis
By David Zierler, Director of the Caltech Heritage Project
January 28, 2022
DAVID ZIERLER: This is David Zierler, Director of the Caltech Heritage Project. It is Friday, January 28, 2022. I'm delighted to be here with Professor Isaac Kim. Isaac, it's great to be with you. Thank you for joining me today.
ISAAC KIM: Thank you for having me.
ZIERLER: To start, would you tell me your current title and institutional affiliation?
KIM: I'm an Assistant Professor of Computer Science at UC Davis.
ZIERLER: What's happening in quantum information at UC Davis? Is there a research group there? Were you brought in to start something? Or was something already entrained before you got there?
KIM: I would say it's a mix. At UC Davis, there was an old-timer in quantum computing, Greg Kuperberg, who was famous for his algorithm. Back in the 1990s, there weren't too many people who had a quantum algorithm under their name, and Greg was one of the few who did. There was an effort, but they're trying to grow in various departments. I guess I'm one of the cohorts who are going to lead that.
ZIERLER: I'm curious, is there any synergy at UC Davis between quantum information and the kind of cosmology that somebody like Andreas Albrecht is doing?
KIM: I cannot say much on the cosmology front, but there are faculties in the physics department, like Mukund Rangamani or Veronika Hubeny, who work on these topics like holography or quantum entanglement. Maybe their efforts do not yet address the cosmology of our universe, but they found some connection between these more traditional branches of theoretical physics and quantum information theory.
ZIERLER: Just a broad understanding, how do you see your intellectual approach and areas of expertise as contributing more broadly to some of the biggest questions in quantum information circa 2022?
KIM: I view quantum information theory as a tool that can ask incisive questions about many branches of physics. The way I think about this is that there are old and important problems in physics, such as simulating strongly interacting quantum many-body systems and understanding the nature of space time. In many of these problems, there are problems, which are hard to solve, obviously, which is why they were so difficult to make progress on. But with these new tools, people are beginning to develop new languages that can lead to more concrete questions. Quantum information seems to be a good language to do that.
ZIERLER: Do you see what you're doing as specifically contributing to the broader effort to build a scalable quantum computer?
KIM: Yeah, definitely. Actually, that's one of the things I first worked on as a grad student at Caltech, developing what's known as quantum error-correcting codes. These are methods to reduce the effect of noise using relatively noisy qubits to make what's called a logical qubit, which consists of many physical qubits but with a much lower error rate. This is one thing I always think about, how to make this more efficient, how to build a fault-tolerant quantum computer in a more robust way. I even ask questions like, "Can we design quantum algorithms so that they have inherent robustness against noise?"
ZIERLER: This has been a fun question to ask because I get as many different answers as there are people I ask. Right now, just on a definitional basis, do we have a quantum computer, however you might define that?
KIM: In my mind, yes. I'm guessing some people might have said no. [Laugh]
ZIERLER: That's right.
KIM: In my mind, yes, because there are devices capable of doing some operations that are difficult to simulate classically. With that said, they're not quite useful yet, but we're trying to make them useful.
ZIERLER: What are the benchmarks you would be looking for to determine that we have built a useful quantum computer, however you define usefulness?
KIM: I think coming from a science background, in my mind, a useful application would be some scientific application, which amounts to solving a scientific problem we had before but couldn't solve instead of coming up with a new problem that somehow suits the quantum computer better. There are canonical problems like this. There's a model called, for instance, Fermi-Hubbard model. It's a very simple model to write down, but computing its properties at its lowest energy state has been very difficult. If we can demonstrate that the result of the computation coming from the quantum computer is more accurate than from the classical methods, I think this would be a pretty conclusive demonstration. People believe this model is rich enough to understand high-temperature superconductivity, so it's not just of an academic interest, this problem.
ZIERLER: I'll note that your answer emphasizes that utility with quantum computers will first be available to help resolve problems in physics and not necessary commercial or economic utility.
KIM: That's my expectation. I think there should be several steps, and the first step should be scientific. If I look back at the history of science, usually it starts with some very simple applications, then it becomes a commercial application, and I don't see why this should be any different.
ZIERLER: For you or at UC Davis, are there industrial partners supporting this research the way that you see Google at Santa Barbara or Amazon at Caltech?
KIM: We don't have any at that level of investment, but I do have some working relationships with these companies in case I need to implement an idea to study these systems. I don't have a quantum computer, so I need somebody's help. For that, I'll need to contact my friends at these companies.
ZIERLER: To the extent that these companies are using different methods to build a quantum computer–Microsoft's and Amazon's approaches are very different, obviously–is there one that strikes you as more plausible? Or is it even as if they're using different methods because their end goals are not the same, they're not looking to build the same kind of quantum computer?
KIM: I think long-term, their end goals are building a fault-tolerant quantum computer, so in that sense, I think their end goal is the same. But I guess they have a different opinion on what would be the most promising technology. The way I see it, I think ions are probably the safest bet when it comes to building a fault-tolerant quantum computer because there does not seem to be a huge difficulty in scaling up the system, and they have shown good control over these ion qubits. But that said, they're kind of slow compared to the other technologies, and that is one thing that worries me. Even if you build a fault-tolerant quantum computer, if you can only perform one operation per second, which is an exaggeration, it's not going to be particularly useful probably. The speed is a nontrivial issue.
ZIERLER: To the extent that these companies are in a race to build a quantum computer, is there any one company or approach that you think is going to win, going to get there fastest?
KIM: This is a much more difficult question. That said, I would say, at this point, it's not even clear if anybody would for sure be able to build a fault-tolerant quantum computer. I'm somewhat biased, but I worked at this company called PsiQuantum, and I think they have a good chance of building a fault-tolerant quantum computer. Their approach is very different. Their mantra is that instead of making a small number of qubits very good and scale that up, you have a scalable procedure already and make that as quantum as possible. I think for various engineering considerations, this is a good approach. I like their chances. But it's a very difficult question.
ZIERLER: Given the fact that quantum information the field is roughly 25 years old, there's tremendous brain power that's being poured into this effort, not to mention the hundreds of millions of dollars from corporations and the federal government, we don't have a quantum computer yet. Why is it so difficult?
KIM: Simply, it's a problem of error rate. Every time you perform a computation, there's a chance that something bad might happen. We call that chance an error rate. Currently, the error rate across very platforms ranges between, let's say, 1% and 0.1%. You might think that's a pretty good number, but if you think about classical computers, the probability that something might happen is ridiculously small, 10 to the -15 or so. You wouldn't even consider that. Ultimately, we're limited by the number of operations that we can perform until the noise overwhelms the system. I think that's the major problem right now.
ZIERLER: Maybe it's a crazy question, but do you need a quantum computer to build a quantum computer?
KIM: I've seen some people saying that. As of now, I'm not quite convinced of that yet because if you look at the existing technologies that people are trying to use to build a quantum computer, like superconducting qubits, ion qubits, cold atoms, topological qubits like photons, we have a very good understanding of the physics behind these systems. Usually, the approach is to build a small enough gadget, which is good enough. For predicting the properties of that gadget, theoretically, I don't see a fundamental difficulty. I'm not quite buying that yet.
ZIERLER: Let's go into nomenclature. Obviously, as you said, the goal is a fault-tolerant quantum computer. What does fault-tolerant mean in this context?
KIM: Imagine anything that you can possibly think can go wrong can happen.
ZIERLER: It's the Murphy's Law of quantum computing, basically.
KIM: Yes. And even then, you'd like to perform quantum computation in a reliable way. That's what fault-tolerant quantum computing is about.
ZIERLER: What are some of the theoretical walls to break through to achieve that?
KIM: We know that, in principle, we can build a fault-tolerant computer. There doesn't seem to be any law of physics that prevents us from doing that. We even have various ways to build a fault-tolerant computer. But in order for the fault-tolerant computer to work, your device has to be sufficiently good. I already talked about this error rate. If the error rate is, let's say, 50%, you do something completely crazy. Clearly, no matter how smart you are, if you have these completely noisy systems, you won't be able to utilize that to build a fault-tolerant quantum computer. There has to be some point below which, once you scale up the system, you can build a fault-tolerant computer, and that number is called the threshold. If we can go below the threshold, which we can these days, then once you scale up the system, you can make the error rate arbitrarily small, which is what we're aiming for. But right now, we're barely below the threshold. While it's true that you can build a good quantum computer by scaling up the system, the overhead needed in making this error suppression can be quite enormous. We call the ratio of the logical qubit to physical qubit the overhead. The big problem is, the overhead is quite enormous at this point, and we need to find a new way to reduce it.
ZIERLER: What are operator extensions?
KIM: Strong subadditivity is one of the fundamental inequalities in quantum information theory. It's a statement about some numbers. You have some numbers, and these are larger or equal to zeroes.
ZIERLER: A term that's very familiar in physics, of course, is entropy. How does entropy apply to quantum information?
KIM: It used to be that entropy was an object that appeared in statistical mechanics in physics. It still does, but entropy in quantum information has various meanings. One of the main contributions of quantum information to physics is that people realized entropy has what's known as operational meaning. Imagine you have some task. Let's say, for instance, I have a quantum system with some entropy of 0.5. What does that mean? You can say that the amount of uncertainty is 0.5. You can say various things, but these are somewhat hand-wave-y in the sense that you can't exactly say what that number means. But in quantum information theory, people found that these numbers are exactly equal to some operationally defined quantity. For example, let's say we have a quantum system with entropy of 0.5. That means that if you have multiple copies of these things and try to extract entanglement, the rate at which you can extract that entanglement is roughly 0.5 if you have this system. This is a clearly defined operational task, and I think you can see various versions of this in quantum information theory.
ZIERLER: We all appreciate the importance of entanglement. In your paper in 2013, you modified that with long-range entanglement. What are the parameters for entanglement to be long-range?
KIM: Long-range entanglement is something more subtle than entanglement. At its most elementary level, entanglement is about a correlation between two quantum systems that cannot be replicated classically. But in these long-range-entangled systems, which is the term we use for certain many-body systems–if you look at correlations between two particles, in fact, you can quickly see that there's no correlation between them in any form. Only if we consider multiple parties of this many-body system do we find that there's a quantum form of correlation between them. That's somehow responsible for many of the fascinating properties we discover in these systems, for instance, for the prospects of using it for storing quantum information or processing quantum information in a robust way.
ZIERLER: Moving to your paper in 2020, what is the quantum marginal problem and its relation to entropy scaling laws?
KIM: Quantum marginal problem is a very simple problem. Suppose I have a large system consisting of many elementary degrees of freedom. You can think of them as electrons, spins, or whatever. But suppose instead of knowing the global information, I only know the local parts individually. I have this collection of particles here, five, a collection of particles here, another five. If I know this local information, can I show that there's a globally consistent state, like a quantum mechanical wave function, that's compatible with this local data I have? It might be hard to think about the motivation behind this question, but it still turns out that for many physical quantities of interest, like energy, magnetization, various order parameters, it suffices to only know the local information, not the global wave function. If you can actually solve this problem efficiently, then you can potentially use it to simulate these physical properties of many-body systems very efficiently. Even to write down this wave function, you would need an exponentially many number of parameters to even write down the wave function. But if you only require this local information, that doesn't scale so badly with the system's size. This problem is extremely difficult, but people found that the solution to this problem can actually come from the entropy scaling law, which was based on John and Alexei's work.
ZIERLER: Even more recently, you had a paper called Chiral Central Charge from a Single Bulk Wave Function. Is there a notion of handedness in quantum information?
KIM: That's a good question. I would say, so far, we haven't thought about this very seriously. But this chirality is something that appears in many physical systems. The short answer is no, but in this paper, we're trying to develop a new quantum information theoretical measure to quantify the chirality.
ZIERLER: I've seen the term robust as it applies to quantum computing and information. You wrote with Brian Swingle in 2017 Robust Entanglement Renormalization on a Noisy Quantum Computer. What is robust entanglement?
KIM: I would say the robust term describes entanglement renormalization, not just entanglement. This paper was about the following question. We already talked about thee quantum computers and noise rates, and we can ask, "In the presence of this realistic noise, what can we do with these devices?" We found that there might be a way to use these noisy devices to simulate physical systems that are currently difficult to simulate. The important point is that there's an existing algorithm to study these systems using classical computers, but that's somewhat costly. The point is, if you replace that difficult part of the computation to the quantum computer, you can speed them up. That's not surprising. The surprising bit is that this algorithm that solves this particular problem happens to be robust against noise.
I think that's pretty interesting because this algorithm was not developed to demonstrate that a quantum computer is more powerful than a classical computer. It's an algorithm that people developed even before people seriously began to try to build a quantum computer, and it just so happened it was robust against experimental noise that's realistic. I think what would be interesting would be to implement these ideas on a real quantum computer to see how accurate of a result we can get. As a matter of fact, only last year, people began to do these experiments at a small scale. This regime is classically intractable, but these devices are getting better. I think we'll see an uptick in these kinds of experiments.
ZIERLER: There have been advances among the experimentalists that even among the theorists, condensed-matter theorists are becoming more a part of the equation right now. Is your sense that in order for breakthroughs to happen, there needs to be greater interaction within the discipline, both on the theoretical and experimental side and among the kinds of theorists who are approaching the issue?
KIM: I think so. I would say there's already quite a bit of interaction between theoretical quantum information scientists and theoretical condensed matter physicists. Maybe what's a bit lacking is an interaction between the theorists and the experimentalists. I think we're at a stage where these experiments are new and interesting. Classically intractable experiments are on the horizon, but these devices are kind of still noisy and not perfect. In my experience, one often needs to go into the weeds and do some not-so-glorious little optimizations here and there. I think this process of getting your hands dirty is something more theorists should do.
ZIERLER: In thinking about collaborations, I don't mean to be American-centric in asking these kinds of questions. Are you involved in research endeavors internationally? Are you following what's going on in China, South Korea, Australia, Europe?
KIM: I do have some colleagues in these places. I was in Australia briefly, so I still have my colleagues there. But I would say this is more at an individual level, not a concentrated effort. Just working on one project at a time.
ZIERLER: Let's go back to undergraduate days at MIT. First question there, you were an undergraduate, the IQI was already underway. Were you interested in quantum information? Were you exposed to that at all as an undergraduate?
KIM: I got very lucky. I got to MIT, and my advisor was Isaac Chuang, who wrote this famous Nielsen and Chuang textbook for quantum computation. That was just pure luck. I didn't ask for that. I was exposed to that at a fairly early age. Ike told me about these things. Peter Shor was also at MIT. They taught classes. I took the class, did some projects, and kind of liked it. When it became time to graduate, I talked to Ike and said, "I kind of like this. Where do you think I should apply?" He gave me a couple places. One of them was Caltech. That's how I came to Caltech.
ZIERLER: You were a double major in physics and math at MIT?
KIM: That's right.
ZIERLER: What would you say was more the home department for people who were pursuing quantum information stuff at that point?
KIM: Good question. This has always been a difficult question.
ZIERLER: I've heard the idea that in the early days, quantum information was kind of homeless. The physicists thought you belonged in math, and the mathematicians thought you belonged in physics.
KIM: Right. Also, computer science. Personally, my home department is physics. After all, I went to a physics grad school. But that said, except for a few places, like MIT or Caltech, I think people view computer science as the home department. At least, that's the impression I got. Physicists are still not accustomed to theoretical computer science work. It was very uncommon. I did take some theoretical computer science courses when I was an undergrad, but that was mostly because I was interested in quantum computation, and most of my friends did not take such courses.
ZIERLER: Besides the advice you got from Ike, did you read any articles from Preskill or Kitaev on what was going on at that point?
KIM: Back then, I just had no idea. [Laugh] I didn't even know who they were. I just asked Ike, and that's what he told me.
ZIERLER: What were your initial impressions when you got to Caltech?
KIM: I didn't get to meet Alexei, but I got to meet John. He seemed to be very knowledgeable about various things. When John talks, he speaks with a lot of essence. I think he uses his sentences very sparingly. At that point, that was the only impression I got, that he seemed to know a lot of things, and Caltech seemed like an interesting place to work.
ZIERLER: Did you jump into IQI immediately? Was that the plan from the beginning?
KIM: To some degree, but I still had some interest in other fields. Mostly string theory. I was a physics kid growing up, and string theory was kind of big. Caltech has a very good group of high-energy physicists, so I thought a little bit about that and even took courses. But at the end, I liked the quantum information better.
ZIERLER: What was the research culture like at the IQI? How would students interact with post-docs, with professors? How did they come about deciding what projects to work on?
KIM: There were many ways in which this could work. There were lots of post-docs. Most of the time, post-docs had their own ideas, and some students worked with post-docs by just asking questions. Some of them asked John for projects. John usually has a couple of projects in mind. For me, that's how I first started.
ZIERLER: Do you have a sense of what John was working on when you connected with him?
KIM: A little bit. I think he was primarily interested in quantum error correction back then. Probably still these days. If it's anything about understanding the fundamental limitation of quantum error correction and how far we can push it, I think it's something he would've been very interested in.
ZIERLER: Was there any partnership or collaboration between the department of physics or the physics option and IQI? In other words, would you get some guidance on the kinds of classes to take that would be relevant specifically within a quantum information environment?
KIM: I wouldn't say so much guidance, but there were very obvious choices. There were quantum information choices, courses taught by John and Alexei. It was very natural to take that. At Caltech, there are some courseworks that are expected of you, such as grad school quantum field theory, I think I took even general relativity, and there's no way around this. I think it's just generally good training for physicists.
ZIERLER: What were the big ideas in IQI once you got your bearings? What was there to choose from of all the things for you to work on?
KIM: Back then, we were interested in this question of self-correcting quantum memory. This is a hypothetical system in which somehow, the underlying dynamics of the system protect your quantum information against the environment. It almost sounds too good to be true because we don't have that. The thing is, if we go to four spatial dimensions, we know that such a thing is possible. I think back then, the big question was, can we do it in the real world, in three dimensions? That led to many interesting discussions and even some progress led by my fellow grad students, like Jeongwan Haah. He came up with a code, which is now called Haah's code. This is something I also worked on, but I didn't find as much nice a model.
ZIERLER: Would you say that the overall motivations at the IQI were more geared towards fundamental research? Or was there a specific endeavor toward applications, moving one step closer to actually building a quantum computer?
KIM: I think back then, it was definitely more fundamental. I remember an event after I graduated in 2014. This was when John Martinis, who was at UCSB then, announced they had built a superconducting qubit device with an error rate below the threshold of what's known as the surface code. That was the point at which I thought, "Wow, we may be able to actually build these things in real life." But before that, in my mind, quantum information and computation were theoretical and experimental disciplines that in the future, at some point, would maybe pan out, but we were still at the fundamental science stage. Things change very rapidly.
ZIERLER: What was the intellectual process like for you in deciding what to focus on for the dissertation, the thesis research?
KIM: There seems to be a lot of connection between quantum information theory and this discipline called many-body physics or condensed-matter physics. I was interested in understanding these traditional many-body systems from the angle of quantum information theory. I just naturally gravitated toward that, so that's what I wanted to work on for my dissertation.
ZIERLER: Was John your advisor?
ZIERLER: What was the process of getting him to be your advisor? Was it formal or informal?
KIM: Pretty informal. John is a pretty hands-off type of guy, so I could work on my own project if I found it interesting. For the most part, I did my own thing. Once in a while, I asked John for advice. That was the nature of our relationship.
ZIERLER: What would you say were the key findings or conclusions of your thesis research?
KIM: I think my thesis research was more of a first step toward a larger program in which we tried to "solve" many-body physics problems using quantum information theoretical tools. Back then, it was more like I developed various tools that could be useful in this direction, but only after I graduated did these tools become actually useful in finding new physics. I viewed it as a steppingstone toward this program.
ZIERLER: Were you following the developments that led to the IQIM?
KIM: Yeah, I definitely remember that. I remember John being very excited. John's a very levelheaded guy, so it doesn't happen very often. [Laugh] This was when I didn't have a very clear idea on how these fundings worked, so I didn't really understand the importance of this event. But once IQI was upgraded to IQIM, my impression was that there was a lot more collaboration, certainly interaction, between various people working, in one way or another, in quantum theory and condensed matter physics.
ZIERLER: The collaboration is where the matter comes in, where you have people coming together.
KIM: Yeah, I think so.
ZIERLER: Did that influence what you were doing at that time? Was that relevant at all?
KIM: More or less, yes. But I would say even when it was just IQI, this was already happening. John already had post-docs who were also on the matter side of things, so to speak. Being in IQIM solidified this position. But at the scientific level, I would say these things were already happening.
ZIERLER: It's more like they were happening, and the NSF recognized that and wanted it to flourish.
KIM: I think so. At least, this is how I remember it.
ZIERLER: What does that tell us about the progress in the field at that point, the fact that these were collaborations that were happening at that point and not, say, ten years earlier?
KIM: Now that I think about it, I think that was when many of the results that we deem important these days were being discovered at a very fast rate. It was just happening at such a fast pace that I guess, for me, I didn't realize that this was rare.
ZIERLER: The theorists were really starting to pay attention to experimentation. The data was coming in and changing the kinds of questions or answers the theorists were after.
KIM: There are a couple things I remember. I guess being a theorist, I'm a little biased, but one thing I do remember is, when Jason Alicea, Gil Refael, this gang of people proposed a very nice way to build these topological qubits using nanowires. Actually, this is now Microsoft's approach to building a fault-tolerant quantum computer. I remember this being quite a big deal as a grad student when I was there. Before then, Alexei had these ideas of building a fault-tolerant quantum computer using what's called topological qubits, but in practice, it was somewhat difficult to make those materials. This new proposal by Alicea et al, the picture they painted was pretty compelling. I think important events like that were happening. Since then, there have been a lot of experiments that have tried to realize this vision. I think this may have been the beginning of all that.
ZIERLER: Besides John, who was on your thesis committee?
KIM: I think it was Alexei Kitaev and Gil Refael.
ZIERLER: Anything memorable from the oral defense?
KIM: I think it was pretty uneventful, so I don't think I have any good stories there. [Laugh]
ZIERLER: Uneventful defenses are also good. What were your options after you defended? What kind of post-docs were you looking at?
KIM: I ended up going to the Perimeter Institute in Canada.
ZIERLER: Who was there at that point?
KIM: Back then, there was Daniel Gottesman in quantum information, also a Caltech alum. There were also a couple people I talked to a lot. This was when Xiao-Gang Wen, who's now at MIT, was there. Also, Dmitry Abanin, who's now at the University of Geneva, was there. Also, down the road from the Perimeter Institute, there's IQC at the University of Waterloo, which is also a big quantum information hub, so it was a pretty good environment.
ZIERLER: What topics did you want to take on as a post-doc? What did you want to expand upon from graduate school? What were new projects you weren't thinking about before?
KIM: This was when Dmitry Abanin was working on what's known as many-body localization. This is a phenomena in which you have a large quantum many-body system, and the system's conventional wisdom would say that any system will eventually thermalize, but these are kinds of systems that never thermalize, in some sense, so it's a very weird system. I remember Dmitry giving a talk about this while I was at Caltech and he was visiting. I thought this was pretty fascinating. Since he was there, I thought it'd be a good opportunity to work on that. That's what I'd been mostly working on there.
ZIERLER: Was it a similar research culture to IQI?
KIM: I think it was very similar, yeah. There were people in quantum information, condensed-matter physics, a lot of discussions, so fairly similar.
ZIERLER: As a post-doc, were you working with graduate students at Perimeter?
KIM: There, it was kind of a unique situation because at the Perimeter Institute, there are no grad students. There are, technically speaking, but that's not the main goal. It's a research institution mostly consisting of faculty and post-docs, so I didn't have much time to talk to grad students.
ZIERLER: To get back to the cosmology/astrophysics, what kind of collaboration was happening at that point at Perimeter among quantum information people?
KIM: This is not something I worked on back then, but I would say between quantum information and high-energy physics, maybe not cosmology, there was a great interest in using quantum information tools to better understand this holographic duality, so to speak. I believe this is still an ongoing effort. I guess a surprise is that many of the very natural quantities that are defined in quantum information as soon as very nice interpretation on the gravitational side of the story, this naturally led to a lot of research activities. There were several people, like Rob Myers at Perimeter Institute, who did a lot of work in this direction.
ZIERLER: I was thinking along the lines of black hole information or Hawking radiation.
KIM: That reminds me. This is not my primary research effort, but I did have one paper. This was because of my Caltech years, actually. In 2012, there was this paper that came out of UC Santa Barbara known as the firewall paper. I remember John talking in the group meeting, and he said this could actually be a really big deal. I've almost never heard him say that. [Laugh] Back then, I had no idea what this paper was about, but it stuck in my mind that it could really be an important topic. I think five or six years later, I ended up writing a paper about that with John.
ZIERLER: Tell me about the next opportunity at IBM. First of all, did you have an appreciation institutionally that at a place like IBM, people like Charlie Bennett were talking about this perhaps longer than anybody else?
KIM: Yeah, that was definitely the main reason. There were also guys like Sergey Bravyi, who's done a lot of very important work. Clearly, a great scientist. It seemed like a pretty good place to work on. Indeed, it was an eye-opening experience for me because before then, I didn't have a chance to talk to experimentalists in great detail about what their capabilities were. I somehow thought everything I wrote down theoretically would be possible in principle. Only after talking to the experimentalists did I realize that, "OK, there are things that can be done in principle, but they require a lot of effort." That really led me to think about, "What can you actually do with these devices we have today?" instead of looking at a longer-term horizon.
ZIERLER: Thinking about devices in the here and now, to what extent is that related to IBM being, to some degree, an industrial environment and not purely an academic environment?
KIM: I think the focuses might be a little bit different. In academia, you'd be more inclined to, let's say, develop a new scientific way to make a qubit as opposed to making your existing qubit better using the same technology but with better engineering. IBM actually does both of them, developing new qubits as a pure scientific research, but also this engineering side. But I guess the focus is more on building a better qubit at a larger scale. It's an industrial entity, so this is completely understandable. In fact, I think it's complementary to academia because only after a huge amount of engineering efforts do you really get to understand the ultimate limit of the scientific approach. I think that was a valuable lesson for me.
ZIERLER: I'm curious at IBM if you gained a window into some of the engineering challenges of building a quantum computer.
KIM: Definitely. For instance, before I went there, I somehow thought that you could scale these systems up very easily, which is not totally untrue, but when you build these larger-scale systems, the yield of making a good qubit can be a problem. Let's say you somehow make a 1,000-qubit device. Not all of them may be perfectly good. The engineering process will not be perfect. That's also important. You need to actually put all these things into a cryogenic environment and control all of them while ensuring that the temperature's low enough. This is a nontrivial challenge.
ZIERLER: Moving onto your third post-doc, it raises a question about the academic job market at that point. Were you applying for tenure-track positions, and there was not much happening at that point? Or did you want another post-doc to continue gaining expertise?
KIM: While I did like my experience at IBM, I think I wanted to do more academic work. Back then, Patrick, who was my post-doc advisor, reached out to me, and it seemed like a good environment, so that's one reason I went there. The academic job market back then, there weren't too many places. At least when I look at the publicly available positions on quantum computing or quantum information theory, I don't remember much, frankly. It's very different these days because there are more positions available for faculty, but back then, it was less common.
ZIERLER: As a Simons Fellow, did you have any connection with either the Simons Foundation or the Flatiron Institute?
KIM: Not with the Flatiron Institute. That's more on the condensed matter side of things. I was actually an It From Qubit fellow. It From Qubit is this program in which they're trying to use quantum information theory to understand the fundamental nature of space time. For that, I was involved a bit with this.
ZIERLER: I'm curious, at Stanford, being in Palo Alto and Silicon Valley, did you get the sense that there were different things happening at Stanford than were happening at Caltech or Perimeter, for example?
KIM: Yeah, definitely. One thing that's definitely missing at Caltech is the startup culture of Silicon Valley. I got to meet with many startup founders and VCs. This was the time when VCs were starting to get interested in quantum computing. Startups you may hear about like Rigetti, IonQ, PsiQuantum were starting then. Eventually, I went to PsiQuantum, which was in the same area. That was one of the influences that led me in that direction.
ZIERLER: Tell me about PsiQuantum. What were they doing? Were they one of the VCs you were exposed to at Stanford?
KIM: When I moved to Stanford, PsiQuantum folks wanted to hire me, but I decided to go to Stanford because I had some academic projects I wanted to do. But I still had a little bit of a working relationship with them, so I did a consulting job with them. Then, later, they wanted to hire me again to basically lead a small team on the application side of things. That's when I decided to move there because it seemed like I could actually hire people and lead the direction in a way that I think is important. That seemed like a pretty appealing opportunity for me.
ZIERLER: What were the academic problems that convinced you initially to go to Stanford first?
KIM: I worked on this thing called the quantum marginal problem. This was probably one of the most challenging problems I ever worked on. It took many, many years. When I was at IBM, I knew that I had something. But if I didn't work on it for the next few years, it was such a difficult problem that I felt like it would just be lost in science history. I felt like I needed to finish it and make it in a publishable form so people could at least understand it if it became important.
ZIERLER: And that wasn't going to happen if you went right to the company.
KIM: Probably not.
ZIERLER: An overall question. Given the fact that behemoths like Google, Microsoft, and Amazon have a ways to go in quantum computing, what can a tiny startup offer?
KIM: That's a good question. These hardware companies actually get quite a bit of funding, so even though they're startups, they're not, technically speaking, tiny. Insofar as you can get these kinds of funding, I think it can still go a long way. Of course, the main question is, how much is enough? This is a difficult question that I don't know the answer to, but so far, they've been doing pretty well. We'll see what happens.
ZIERLER: Early in our conversation, you said that PsiQuantum's doing some pretty exciting stuff, that they might actually get to some major breakthroughs. What's their approach? What are they offering that others might not be?
KIM: They're based on a photonics approach. The basic idea–this is all in the public domain, by the way–is that they will create these entangled states of photons consisting of, let's say, ten qubits. Imagine you have this little device that spits out these entangled photons all the time. What you can show is, if you can generate them at a fast rate, by performing some entangling measurements between them, you can perform universal fault-tolerant computation. What I like about their approach is this maybe very simple idea, that once you build a device that's capable of doing a fixed thing like spitting out these entangled photons, it's good enough for a fault-tolerant computer. The question is, can you actually build that? We don't know. We'll see what happens. But it somehow seems like a very nice approach to this problem. Instead of having, let's say, millions of qubits arranged on a chip, it seems like once you're past a certain point, you're pretty much set.
ZIERLER: When the opportunity at the University of Sydney came along, was COVID already a problem by the time you got to Australia?
ZIERLER: You got in just before.
KIM: Well, COVID started one month after I got there.
ZIERLER: This is when there was already an outbreak in China, but it had not become an international problem yet.
KIM: It did not seem that serious back then.
ZIERLER: Of course, Australia has taken some extreme measures. What was it like living in the pandemic in Australia during that year and a half?
KIM: If you're inside Australia, I think, for the most part, you're pretty good. It wasn't a bad environment. The thing is, most of my family and friends are outside of Australia, so it was somewhat difficult for me. Before COVID, Australian schools and government were quite lenient on travel. That's one of the reasons I thought it would be OK to go there. Scientifically, they have great people there, so that part is excellent. The only question was whether I could still keep in touch with people in the States, for example. This didn't seem like a huge issue before COVID. Only after COVID did it become a serious issue.
ZIERLER: Was that a tenure-track appointment? What does lecturer mean in the Australian system?
KIM: It's like a UK system, so it's slightly different. Technically speaking, there's no tenure-track program in Australia like in the UK. It's difficult to compare.
ZIERLER: It was more than a post-doc, though, you were on the faculty.
KIM: Yeah, I was on the faculty.
ZIERLER: What was going on at the University of Sydney? What was some of the research on quantum information?
KIM: There are people like Steve Bartlett, Andrew Doherty. They worked in various fields in quantum information, quantum error correction. There were also faculty like Mike Biercuk, David Reilly. They have quite a bit of faculty already working there. Also, they were hiring other young faculty, like Arne Grimsmo. Overall, for a single school, it's hard to have that many faculty working on quantum information. I guess these days, there are more schools that can claim that, but back then, there weren't many.
ZIERLER: Was this largely a remote life for you in Australia? Were you going into school, were you interacting with students and colleagues in person?
KIM: I was. For class, I had to do it remotely, but I could still talk to people in person, which was nice.
ZIERLER: Were there companies like in the United States that were supporting Australian research in quantum information?
KIM: There's a startup called Q+Ctrl, which was founded by Mike Biercuk, my former colleague. I'm sure there are other smaller companies, but not in the Sydney area.
ZIERLER: To the extent that there's always a cultural component to where research happens, was there a uniquely Australian approach to quantum information that you might not have seen in the United States or Canada?
KIM: I feel like people were more relaxed and down to chat with each other compared to, let's say, my Caltech and Canada years.
ZIERLER: A little more laid back in Australia.
KIM: Yeah, it's quite nice there. It's a nice city. You can have lunch and a nice coffee. I think it's actually helpful.
ZIERLER: What was some of your key work in Sydney?
KIM: This was when I really seriously began to work on the interplay between quantum many-body physics and quantum information theory. I already told you that during my grad school years, I did some technical work that could be useful for this. This was when I basically found my colleagues who work in the same direction, guys like Kohtaro Kato, who's also a former Caltech alum. We didn't overlap, but he was there. Also, Bowen Shi, a post-doc at UCSD. This was when this program really came into fruition. The thing is, before this period, my work on quantum information theory and quantum many-body physics stopped at reproducing the results that people already knew. But in this period, we actually began to find new things that people didn't know before. It really gave me confidence that something new was happening.
ZIERLER: Tell me about the opportunity at UC Davis. How did that come about? Were you specifically looking for faculty positions back in the United States?
KIM: I was looking for faculty positions back in the States because of personal life issues.
ZIERLER: Most of your family is in the States?
KIM: My parents are in Korea, but my sister and most of my friends are here. I was looking for opportunities in the States, and there were more opportunities than I remembered compared to my Stanford years, for example, which was good. This was one of the programs that was funded by NSF. NSF had this program of funding some schools for hiring quantum computing faculty. UC Davis had one such opportunity. It seemed like a good fit.
ZIERLER: Did you see that as an opportunity, joining the faculty, to take on new positions, to bring what you were doing from Sydney?
KIM: I think primarily. UC Davis and University of Sydney were both scientifically interesting, but I think I probably wouldn't have moved unless there were personal reasons. There's a lot of overhead in moving. Even if scientifically, it makes sense, it can be challenging. It's a challenging life decision. That said, UC Davis seemed like a pretty good environment. Greg Kuperberg was there. I had some work I thought would benefit from being in the computer science department as opposed to the physics department. There was a large group of people, like Veronika Hubeny, Mukund Rangamani, Bruno Nachtergaele. These are all people I at least am acquainted with and know their work very well. It seemed like a stimulating environment.
ZIERLER: To bring the conversation right up to the present, was one of the plans when you got to UC Davis to build up your own research group, to recruit graduate students and post-docs?
KIM: That's definitely the plan.
ZIERLER: How's that going? How far along are you right now?
KIM: So far, I'm just talking to the students who are already here as opposed to recruiting new students. This is mostly because I came in July, so I was a little short on time to advertise before the grad school application cycle. This year, I'll start to advertise more.
ZIERLER: In academic terms, you're still brand new.
ZIERLER: What kinds of graduate students are you looking for? What are the things where this is the overall approach of your research group, these are the kinds of interests you're hoping your graduate students will have?
KIM: I guess the big question I'm trying to solve is to use these quantum computers in understanding quantum entanglement to efficiently simulate these many-body quantum systems. I think there are various ways in which the students can contribute. One is more mathematical. Someone who would like to understand the fundamental nature of these quantum correlations in these many-body systems. At the opposite extreme, there will also be a large amount of practical work using these somewhat clunky near-term quantum computers and a large amount of coding. I think any student who's sufficiently talented will be able to contribute something substantial.
ZIERLER: What about on the undergraduate side? Are you getting the sense that undergraduates, particularly since so many are interested in computer science, are interested in quantum information, about where quantum computation might be for their careers?
KIM: There's certainly a lot of interest. At the very least, a lot of curiosity. I'm actually going to teach a quantum computing course for undergrads, and two days ago, I gave a short, 15-minute presentation on what the course is going to be. The students seemed to be quite interested. They are interested in the job prospect as well because it's a new field. The interest is there.
ZIERLER: In an undergraduate course, Intro to Quantum Computing, or whatever you might call it, where is the field now in terms of what you'd expect undergraduates to come with? What would be the pre-reqs for that? Linear algebra? What kind of physics background would you need to understand, "Here's where we are with quantum computing circa 2022"?
KIM: I think it depends on the department. If you're in the physics department, I think you probably want to have quantum mechanics as a prerequisite. If you're not in the physics department, like computer science or electrical engineering, I think there's a way to teach quantum computing that just requires linear algebra.
ZIERLER: Now that we've worked right up to the present, for the last part of our talk, I just want to ask a few broadly retrospective questions, then we'll end looking to the future. As you look over your trajectory, all the institutes you've been a part of, all your collaborators, do you see a particular trajectory along a line of progress? In other words, you were only able to work on this because you worked on that previously. Do you see a narrative to what you've been doing and what that might say about the field writ large?
KIM: I think looking back, my research was definitely influenced by the people who were around me. At least when I was at Caltech, and during my post-doc years, it seemed like there were a lot of interesting things happening around quantum information theory and quantum many-body physics, so it was a very natural thing for me to think about. Also, this information paradox and quantum information theory is another topic that seemed interesting and many people thought was important. I don't think I would've worked on these problems had I not had that kind of experience. Looking forward, I think something similar. I'm also thinking more about what we can do with these real experiments we're going to be able to do in the near term. My gut feeling is that this will probably influence my research a lot.
ZIERLER: As we look to the 25th anniversary if IQI/IQIM. At an institutional level, what do you see as Caltech's contributions as we move further along in quantum information and a scalable quantum computer? Not just in terms of the individuals who have come through the program, but maybe even the intellectual approach or the vision that John Preskill had from the beginning.
KIM: In my mind, the most important contribution of IQI/IQIM is advocating a rigorous scientific approach to this new field. By rigorous, I mean brining the highest level of scientific integrity. I think in a new field like quantum computing, there is a potential danger of making unsubstantiated promises, which can dilute the trustworthiness of the field. I think what IQI succeeded in doing was that John made sure everything that came out of there was based on rigorous mathematics and science. When you do that, even if the field is new and in the early stages, you can firmly say what is and is not possible. Showing that something remarkable like building a fault-tolerant quantum computer is possible in principle under a certain set of assumptions was a very important scientific approach in terms of solidifying quantum computing and quantum information as a legitimate field. In my mind, that's the most important contribution of IQI.
ZIERLER: Finally, for you, looking to the future, what have you learned about progress in the field and the trajectory of things you've worked on that might influence or help you establish a framework for what to work on next.
KIM: I've been thinking a lot about what constitutes an important problem. I think it's still something I'm working on. I think I'd like to choose something that, if the research direction succeeds, it requires us to change our conventional way of thinking that we've had for years because that suggests there's something fundamental we didn't know about a certain discipline. I always think about several instances in which John or Alexei said something was important. They did say that, but only very rarely. I do always wonder how they came to that conclusion. My feeling at the moment is that it's either something fundamental and simple enough that it can influence the thoughts of many people, or it's something that can drastically improve things. Let's say we have a fundamentally new approach to building a fault-tolerant quantum computer that, in the long run, will outperform our current best methods. That would clearly be of a great scientific interest. These are the types of questions I would like to find and work on in the future.
ZIERLER: But the idea is that the field is exciting, there's no shortage of opportunities to work on. There's plenty to do.
KIM: Oh, there's plenty to do, for sure.
ZIERLER: On that note, it's been great talking with you. I'd like to thank you so much for doing this.
KIM: Thank you for leading me through all of this.