John Martinis

Notice: We are in the process of migrating Oral History Interview metadata to this new version of our website.

During this migration, the following fields associated with interviews may be incomplete: Institutions, Additional Persons, and Subjects. Our Browse Subjects feature is also affected by this migration.

We encourage researchers to utilize the full-text search on this page to navigate our oral histories or to use our catalog to locate oral history interviews by keyword.

Please contact [email protected] with any feedback.

ORAL HISTORIES
John Martinis

Credit: UC Santa Barbara

Interviewed by
David Zierler
Interview date
Location
Video conference
Usage Information and Disclaimer
Disclaimer text

This transcript may not be quoted, reproduced or redistributed in whole or in part by any means except with the written permission of the American Institute of Physics.

This transcript is based on a tape-recorded interview deposited at the Center for History of Physics of the American Institute of Physics. The AIP's interviews have generally been transcribed from tape, edited by the interviewer for clarity, and then further edited by the interviewee. If this interview is important to you, you should consult earlier versions of the transcript or listen to the original tape. For many interviews, the AIP retains substantial files with further information about the interviewee and the interview itself. Please contact us for information about accessing these materials.

Please bear in mind that: 1) This material is a transcript of the spoken word rather than a literary product; 2) An interview must be read with the awareness that different people's memories about an event will often differ, and that memories can change with time for many reasons including subsequent experiences, interactions with others, and one's feelings about an event. Disclaimer: This transcript was scanned from a typescript, introducing occasional spelling errors. The original typescript is available.

Preferred citation

In footnotes or endnotes please cite AIP interviews like this:

Interview of John Martinis by David Zierler on May 4, 2021,
Niels Bohr Library & Archives, American Institute of Physics,
College Park, MD USA,
www.aip.org/history-programs/niels-bohr-library/oral-histories/46976

For multiple citations, "AIP" is the preferred abbreviation for the location.

Abstract

Interview with John Martinis, professor of physics at UC Santa Barbara. Martinis gave the interview from Australia, where he was consulting for Silicon Computing following his affiliation with Google’s efforts to build a quantum computer. He surveys the current state of play toward that goal, and explains what applications quantum computing can serve, and how the field is clarifying the technological requirements to achieve a quantum computer. Martinis recounts his childhood in Los Angeles, his early interests in computers, and his undergraduate experience at Berkeley where he gravitated toward experimental physics. He describes his interactions with John Clarke and his motivations to stay at Berkeley for graduate school, where he focused on SQUIDS and was captivated by Tony Leggett’s ideas on quantum tunneling. Martinis explains his interest in working with Michel Devoret at Saclay for his postdoctoral research, where there was much excitement over high Tc and YBCO materials. He describes his subsequent work at NIST and his decision to join the faculty at Santa Barbara around the time he became focused on quantum computing. Martinis narrates the technological challenges of building qubits and error correction, and he explains how he got involved with Google and joined his style with its research culture. He describes his role as chief scientist in the collaboration and why his vision and Google’s diverged. Martinis addresses the issue of “hype” in quantum computing. At the end of the interview, Martinis emphasizes the centrality of systems engineering to his research agenda, and he explains why quantum supremacy will demonstrate the need for quantum computing and the limitations of classical computing.

Transcript

Zierler:

Okay. This is David Zierler, oral historian for the American Institute of Physics. It is May 4th, 2021, for me. And it is May 5th, 2021, for Dr. John M. Martinis who is coming to me from Australia. Thanks to the magic of Zoom we can talk in real time halfway around the planet. John, it’s great to see you. Thank you so much for joining me today.

Martinis:

Yes, thank you very much.

Zierler:

John, to start would you please tell me your title and institutional affiliation?

Martinis:

I’m a professor of physics at the University of California, Santa Barbara. Right now, I’m at Silicon Quantum Computing in Australia, in a consulting role. I’ve been here for about eight months now. I also started a small company called Quantala, which is mostly an intellectual property holding company for some ideas I have come up with. It’s also a way to maybe start something bigger if something interesting happens.

Zierler:

Now, do you also retain an affiliation with Google?

Martinis:

No. I left Google about a year ago. I have a student that is working at Google, and he is doing a nice experiment on the effect of cosmic rays on qubits.

Zierler:

Now, if I understand the chronology, you’re in Australia now not because you were there on a consulting trip and got stuck during the pandemic. Eight months ago, you went in the middle of the pandemic. That was already in the cards when you went down.

Martinis:

Yeah. That’s right. After I left Google, Michelle Simmons contacted me and suggested that I could come work there. There was some family business to take care of, but around June I decided, yeah, it’d be great to work with them and learn about silicon qubits. She then arranged with the government to get a special visa for me to work in Australia. In the middle of September, we flew to Australia, and then had to go through the two-week quarantine. That wasn’t so nice, but afterwards it was great that everything was more or less open since there weren’t very many cases. And it’s been kind of normal here for eight months. I’ve really enjoyed working with Michelle, as she and I think quite similarly about physics and quantum computing, and it’s also been nice to learn about semiconductor qubits. It’s also been interesting to talk about leadership and running a lab. So, this has been a great sabbatical for me. My wife and I will be going home in a few weeks, which seems to be good timing since the infection rate is low and we can get vaccinated right away. It’s been nice to escape COVID for these months (laughter).

Zierler:

Now, have you retained responsibilities in Santa Barbara? Have you been teaching remotely?

Martinis:

Well, it turns out that the arrangement with Google had me retaining my position at UC Santa Barbara at ten percent time. After I left Google, I have been figuring out what to do next. The nice thing is UC Santa Barbara has been very accommodating for me in this whole journey. I like teaching and I like Santa Barbara and the university. They’ve been very nice and understanding. I’ve always wanted to build a quantum computer, which is of course why I decided in the first place to move to Google and work there for seven years. So, in the end it didn’t work out for me, so now I am working on setting up some collaborations with various other groups to continue work on quantum computing.

Zierler:

John, before we go back and develop your personal narrative, I’d like to ask a few contemporary questions that I think will punctuate our discussion. The first is of course in this past year-plus, how has your science been affected one way or the other with the mandates of social distancing and remote work? In other words, have you found that absent commuting and being in person that you’ve had more bandwidth to get stuff done? And alternatively, has the inability to be physically connected with your collaborators stymied your research and has prevented more advances than you may have hoped over the past year?

Martinis:

Leaving Google was a much bigger event than COVID for me, although they were connected since part of the reason I left was to lower my stress levels, to be healthier in case I got it. I’m an experimentalist, and I have been running a group for a long time, so this was a big change. And I enjoy working with many people and advising them, and spending my time trying to guide a project and understand the big picture. That’s what I’ve been doing for many years now as a senior scientist, it was the dominant thing in my life, and I don’t have a big group to collaborate with anymore. So, coming to SQC and working with Michelle and her group has been really great. I also have been working on some physics of superconducting qubits on how quasiparticles and cosmic rays affect the qubits. My student was working on this, and I developed some theory on how to holistically describe this, and so I have published a paper on the ideas. I also had an idea on the design of qubits using tapered wiring to lower surface loss, and also published a paper on that topic. The plan next is to work with other groups and help in whatever way I can. With COVID, it actually has been working out fine since everyone is used to video calls.

Zierler:

It was an opportunity.

Martinis:

It was an- you know, not the kind of positive opportunity that you normally look for (laughter).

Zierler:

It’s a pivot. It’s a pivot point.

Martinis:

Yeah. It was a pivot point. I think it’s time for me, for everyone, to move on and do something different. I have been thinking that this is the next stage of my career, and I have an opportunity to try a bunch of new things and ideas. Working at SQC was really interesting for me, and I was glad to have the opportunity to learn new physics.

Zierler:

I’m curious about your perspective with your very unique career trajectory. Coming from the background of academic physics which is very much a basic science atmosphere, what have you learned about navigating the tricky waters administratively, legally, in terms of intellectual property, when you have an idea that has commercial viability and you want to go for a patent? What are some of the big lessons you’ve learned over the course of your time doing these things?

Martinis:

Well, I’ve always been interested in startups, even having taken a business plan course in Colorado years ago. I’ve read books about patents and helped lawyers write them in the past, so I know the structure and core ideas of a patent. And when I had some ideas this year I kind of knew there were some core ideas that could be valuable. I found a patent lawyer through UCSB, and I worked with them to write two patents.

Zierler:

Now, as you well know, the term quantum computing or quantum computer, it’s a buzzword. It’s in the zeitgeist. The public hears these words. So, with that in mind, what is the biggest misconception about what a quantum computer is?

Martinis:

You know, I actually don’t worry so much about the public misconception. What I worry about is that scientists need to talk about a quantum computer in a way that is realistic enough so that we can have long term success. Particularly, I worry about hype. If scientists get funding, for a startup or whatever, it’s great, but the critical step is to build a quantum computer. You need to set up the funding to make progress in the long term, as it’s going take time. One needs funding for a decade or more, so one has to be patient. Also, there’s been hype cycles in the past for quantum computing, and although the field has made a lot of progress recently, it is perhaps a bit bigger this time. I’m trying to inject a dose of realism to the field when I speak about quantum computing, and I think people generally appreciate it. Looking into the future, I don’t know who the technology winners or losers are going be, we just need funding to do the research to find this out.

Zierler:

On the question of being realistic as I’m sure you well know, many of your peers, many people in the field will assert we’re not even sure at this point what quantum computers will be good for. So, my first question there is do you accept the premise of that statement? And either way, scientifically how is it feasible to have an idea to build a quantum computer without knowing necessarily what its end use is going to be?

Martinis:

Right. So, first of all, I think the statement you made is not quite correct in that we do see a definite application right now. And then there’s other applications we’re not so sure of, but would maybe have a much bigger impact, so it’s worth looking at those too. So, let me explain. The application that looks really compelling, Feynman’s original proposal, is computational chemistry, solving the quantum mechanics of chemistry and materials with a computer. This is an important field right now since about fifteen percent of supercomputer time is for this application, and clearly if we can map this quantum problem to a quantum computer, it can be solved much more quickly. Feynman first had this idea in the mid-1980s, but it wasn’t until the mid-2000s that people started understanding how to do the mapping. And then it took ten or fifteen years for people to improve the mapping and make the algorithm compact enough to fit on a quantum computer that eventually one could build.

This was something that the Google theory team was working on when I was there, and they estimated that it would take about a million physical qubits, which is about a thousand logical qubits that are error corrected, to start doing quantum chemistry that’s scientifically interesting and practical. You probably would want to build something larger in the end. For me, this way of thinking is like what you find in the book by Peter Thiel, Zero to One, where he talks about having a definite positive outlook on startup technology. Here, we have a definite application where you know you can build a quantum computer to do something scientifically and technically useful. Even if you need to spend a lot of money to do this, billions of dollars, then practical things will come out of it, say designing better chemical reactions or better batteries.

So, it is worth the effort, we know what algorithm to try, and it seems within reach. That’s a great motivation to move forward. Now looking further into the future, a far bigger application is using quantum for machine learning and artificial intelligence. The basic algorithm for AI is to fit some abstract mathematical function to data that’s out there. Right now, you have to do that with classical computers using huge data sets, so you can think of this as slow learning. The idea is maybe with quantum you can train with much smaller data sets, learning more efficiently or quickly. If you can do this, there are many applications that are useful, such as the travelling salesman problem, airline scheduling, a long list of optimization problems that have huge economic value. The problem is that we don’t yet have an algorithm. Since it took many years to understand the use of deep neural networks for classical AI, invented in a heuristic way, then maybe that is how it will progress for a quantum computer. So, I have an indefinite optimist’s view on this application, that scientists will eventually be able to solve this problem too.

Zierler:

Given that you have such well-formed ideas about quantum computing applications, I’d like to ask a question that is, it really pushes on the boundary between science and philosophy. And that is, you know, right now in the world of experimentation in physics, obviously we’re limited by instrumentation. We’re limited by budgets. We’re limited by all of these things. So, let’s imagine a world, heaven forbid, where LHC is the best that we’ve got. We never get the ILC. We never have a project that operates at high enough energies to find new physics. To see something beyond the Higgs. To find supersymmetry. You name it, right? Can you envision a world in which quantum computing can simulate those nuts and bolts experiments? And on top of that if we find supersymmetry in a simulated environment can quantum computing be so good and so effective that it will be understood as deductive evidence the way that we have only in a world of precomputing and classical computing?

Martinis:

That’s the long-term vision which I think is so interesting about quantum computing. This particular application you’re talking about is along the lines of quantum materials and quantum chemistry. Now, my view is that you have to do the quantum chemistry and quantum materials problem to validate the quantum computer. You need to run these algorithms, write the scientific papers, and then you’re going to get better at learning how to solve these problems. As an experimentalist, it seems that you eventually want to attach this research to the physical world.

Zierler:

Yeah. Yeah.

Martinis:

But you know, maybe in some of these fields, computation will be the best we can do. And the quantum computer will help people understand those problems. And it’ll change the way people think about doing physics, maybe in the same way that science is done differently now because of classical computers.

Zierler:

Right.

Martinis:

But just like classical computers, you have to learn techniques to test things with experiments so that you can trust these answers. It will probably take a while to figure that out.

Zierler:

Well, our discussion has been very forward looking up to this point. Let’s go backward. Let’s start first with your parents. Tell me about them and where they’re from.

Martinis:

My dad was born in Croatia. As it happened, his father left for the United States shortly before the war. And the family left behind were kind of by themselves over the war and it was very, very difficult for them. My dad actually was a child soldier from the age of thirteen to seventeen. So, he had a very difficult childhood, and especially the teenage years, and actually never talked about it much. I think he did quite well for having such a difficult time growing up. After the war, his family was trying to get to the United States. Yugoslavia was beginning to close up at the time because of the communist government. In the end, escaped from Yugoslavia and had security people chasing him to make an example of him. In the old country his family fished. So, when he went to Washington state that is what he did with his father and brother. During the off season he came down to San Pedro in California, where he had relatives, and met my mom. And he was a fisherman in San Pedro for a few years, but that industry was declining, and a neighbor suggested becoming a fireman. He applied and it was a perfect job for my dad as it is very physical and technical, you have to understand how things work.

My mom was a homemaker. The great thing about my dad was that he was always building projects in the garage, and I got to help him out. So, I grew up learning how to build things, so I was able to practically understand physics even as a young child. And when I finally learned math, and especially when I took the high school physics course, it’s like all this made sense because there were formulas behind something I already intuitively understood. I kind of knew what would happen. I was good at physics and loved how it explained how things worked, so I decided to follow it as an undergraduate.

Zierler:

Where did you grow up?

Martinis:

In San Pedro. This is the harbor of Los Angeles.

Zierler:

Mm-hm.

Martinis:

In California.

Zierler:

You went to public schools throughout?

Martinis:

Yes, yes. The public schools I went to were good. At my elementary school, they had a special science center not far from our school, and I was able to take some special classes.

Zierler:

When did you start to get interested in physics specifically on an academic level? Was it high school?

Martinis:

I was always interested in math and science, but it was the physics course in my senior year that got me hooked on it. And I had a very good teacher, who explained physics conceptually, in a way I had not experienced in other classes. Since I understood many of the physics concepts from building things, I was able to understand deeply how the math worked. I also had an interest in computer science at that time, and there were no computers at our high school, but I was able to take some night courses. I remember picking up a book in the library about programming using FORTRAN, and just devoured the concepts since it was so interesting. I was thinking at the time to maybe become a computer scientist, but I enjoyed and liked physics so much, I did that instead. Given my background in understanding physics intuitively, that was more natural for me in the end.

Zierler:

Besides Berkeley, where else did you apply for undergrad?

Martinis:

The place I really wanted to go was Caltech, which you hear about all the time growing up in Los Angeles. I even went to a few evening lectures there during high school. Unfortunately, I was not accepted, I think because my English scores were not that great. Growing up, I was just more interested in building things and understanding how things worked instead of reading books. But in the end, Berkeley was better for me.

Zierler:

Now you got to Berkeley in when, 1976?

Martinis:

Uh, yes. The fall of ’76.

Zierler:

Were there any vestiges of the counterculture of the antiwar movement? Was your sense that that all died down by the time you got there?

Martinis:

Well, what’s funny is my parents were really kind of worried about me going to Berkeley because of that.

Zierler:

Yeah.

Martinis:

One should remember in 1976 these protests were fairly recent history. Fortunately, my cousin had gone there the year before and things had kind of turned out okay. And Berkeley kind had a great reputation for physics and, well, was farther from home and thus good for my independence. I was really focused on physics, and that’s what I spent my time doing. My parents were happy to see that I did well there.

Zierler:

Now was it experimentation all the way? Or were you open to theory as an undergraduate?

Martinis:

Oh, I always felt I was an experimentalist, from the very beginning. I did quite well in the courses, even with the honors courses. But things really took off my third year when I got to the upper division experimental lab, and then the following year when I did research for my senior thesis. I really felt at home at that point and could do it as a career. There was a certain style at Berkeley doing experiments having to do with Ernest Lawrence, which I’ve only appreciated in the past few years by reading the biography of Ernest Lawrence and the radiation laboratory. So, for my senior thesis I built a nuclear physics detector, called a Hadron calorimeter. I was lucky enough to do an experiment with it at the Bevalac, which is the machine where they discovered the antiproton. What I hadn’t realized at the time is that I was basically building an experimental apparatus that one could use for a PhD thesis, although I would have had to understand what I was doing in much more detail. Looking back, it was amazing that they gave me such an interesting and challenging project.

Zierler:

What professors or courses were particularly formative for your intellectual development as an undergraduate?

Martinis:

Well, the freshman and sophomore honors courses that I took were really kind of special when I look back on it. Another professor at UC Santa Barbara took these classes too, and we have talked about how amazing it was. Mark Strovink was the professor for the first two classes who really taught challenging courses with lots of rigor. I liked it, as it was my first experience with university physics. And then for the second year, for electromagnetism, the course was taught by J.D. Jackson, okay? So, you know who he is?

Zierler:

Of course!

Martinis:

He wrote the classic graduate textbook on electromagnetism. It was amazing to learn electromagnetism from the guy who wrote the book on it (laughter). The funniest thing is we didn’t realize it at the time, but he sometimes gave us some exam problems from his book. Okay, the simple ones. And if we would’ve thought to also read his book we would’ve done better. He was just a great professor. Additionally, our textbook was the book written by Purcell, which is famous from the Berkeley physics series. A classic book. And what’s great is because I had such a good foundation from the professor and the book, and electromagnetism has always been a strength for me. So, I enjoyed the rigor and depth of university physics, and did pretty well because I was motivated to study hard at the time.

Zierler:

Now did you have interaction with John Clarke as an undergraduate? And did that affect your decision to stay for your graduate program?

Martinis:

Yes, yes. During my senior year, while I was doing the thesis experiment on nuclear physics, I took a course in condensed matter. I was still deciding what I might want to do. I was thinking about the long-term prospects of nuclear physics, and particle physics was interesting but involved huge groups, so I wasn’t quite sure that direction was for me. John Clarke’s class was wonderful, and I became quite interested in condensed matter physics. and started talking about doing research with him as a graduate student. But the idea that really got me interested was that John was doing experiments on quantum noise and quantum effects in devices, which of course is the precursor of all this quantum computing research.

Zierler:

He was ahead of his time. You recognize this now.

Martinis:

Yeah, yeah. He was doing an interesting experiment measuring quantum noise in superconducting devices, which at the time was a bit surprising to everyone. In my graduate course in quantum mechanics we learned very early on about the Bell inequalities. So, physics was beginning to look at quantum effects in new ways, and this seemed super interesting. From high school, I was always interested in electronics, and I thought it was interesting to build new devices and study them for quantum effects, which was all very new at the time. So, I talked to John about doing this kind of research. I applied and was accepted to several different graduate programs, but this was at a time when the physics graduate programs didn’t really recruit. I had good grades from Berkeley, an NSF fellowship and good GRE scores, but no one recruited me, even to pay for a visit. In thinking about this, I was super interested in working with John Clarke and doing new and interesting physics, so I just made the easy decision to stay at Berkeley. I knew I would fit in and be happy. I guess this was a good choice since this is what I have been doing for my entire career. I really liked John Clarke as an advisor. He is very nice and taught me a lot and had very high standards. I liked his research style and the way he advised his group.

Zierler:

Which is what? What is John’s style like as an advisor?

Martinis:

Well, the most important thing is that he came into the lab and talked with you every day. At the time it was sometimes annoying, but looking back this was probably the single most important thing he did to train me as an experimentalist, okay? (laughter) You know, you need help as a student to talk about what you are doing, every day. He was also very good at writing clear papers and giving well-organized talks. John came from Cambridge, and our scientific lineage thus goes to Pippard and Kapitsa and eventually all the way back to Rutherford. So, I’ve appreciated over the years having this scientific heritage that emphasized how to do physics well.

Zierler:

As you were putting together your thesis topic with John, in what ways was theory serving as a guide at this point in the field? What was going on in theory that was relevant as you were thinking about experiments and observation.

Martinis:

Yeah. So, let me first talk a bit about what happened as a graduate student. So, the first project was to measure the input noise in a SQUID amplifier, which had never been measured before. So, it’s kind of a hard experiment, at least for a new student, okay? At first, I had to build the devices, and fabrication took me about a year just to understand since I had to develop new processes, and frankly I just make lots of mistakes. This was a great starter project that John assigned to me, since I had to work on my experimental skills and understand noise. So, I have always thought carefully about starter projects for all of my students.

During this time, I had gone to a conference and watched the talks by John Clarke and Roger Koch on quantum noise. But what really piqued my interest were theory talks by Anthony Leggett on macroscopic quantum tunneling. His thesis was that we know quantum mechanics works for particles, electrons, and atoms and fundamental particles, but we did not have experimental evidence that it works for macroscopic objects. So, if you had a ball, and threw it against a thin enough wall, then would it tunnel through the wall? It’s an interesting question because of the basic laws of quantum mechanics, and the paradox of Schrödinger’s cat can be thought of as whether such a macroscopic object could be in a superposition of the dead and alive quantum states? Maybe nature just does not allow this to happen, and a macroscopic state is what collapses the measurement into a dead or alive state, and thus you should test it.

Zierler:

Yeah.

Martinis:

So, in the Josephson junction, there is a macroscopic collection of Cooper pairs that tunnels through the junctions, a macroscopic number of Cooper pairs flowing through the device. Thus, superconductors are an ideal system for testing whether macroscopic variables obey quantum mechanics. As a student, it totally fascinated me that one could do such a fundamental experiment, and the results would be really interesting. At the time there were some experiments that were still a bit ambiguous. So, I talked to John about how I would really like to do this experiment and to do it right, as it was worth doing well. Of course, John understood that, but wanted to make sure we would be doing something different and better than what was done before. So, we got to thinking, and the basic idea was to measure the device parameters so that we could do a direct comparison of the experiment to theory, with no adjustable parameters. Does this make sense?

Zierler:

Absolutely.

Martinis:

At the time Michel Devoret came and worked for John Clarke, and he understood how to make and use a dilution refrigerator, which was absolutely critical to run the experiment. Michel is very nice and an incredibly smart physicist, and it was great to have deep conversations every day and to mentor me. So, this experiment really showed me what a good collaboration was all about. For example, very soon we were talking about how to measure the parameters of the experiment, and then I think John Clarke came up with the idea to measure the oscillation frequency. When you drive the system with microwaves that are resonant with the system, you give it energy and will observe a resonant increase in the measured switching rate. What is so interesting is that in the quantum regime, this allows you to observe the quantized energy levels, which is the clearest “smoking gun” evidence that the system is obeying quantum mechanics. So immediately we decided to classically simulate the effect, and since it was the mid 1980’s the fastest way to do it was with analog electronics, using a Josephson junction simulator we had in the lab. So, in a day or two I was able to validate that this idea would work, and then Daniel Esteve in Saclay simulated it on a computer to give more accurate predictions. At this point we knew how to start working on the experiment.

Zierler:

How did you know you had enough to defend? What was the feedback you were looking for in the experiment?

Martinis:

So, we started building the experiment and testing it to see if it made sense. At first, we just simply tested it between four Kelvin and one Kelvin in liquid Helium, just to check if the classical thermal physics was working right. The measured switching rate was predicted to scale in a certain way, and it just didn’t make sense. I think we figured out quickly that the problem was that we were seeing noise from room temperature that was coming down the wiring, which dominated the physics. The key idea was to make low temperature filters so that the only noise that the device was seeing was coming from the device temperature. Using Michel’s experience with nuclear magnetic resonance, we were able to make some Copper powder filters that worked really well, as well as starting to test the chip mount with microwaves to directly measure the amount of filtering.

It was not too complicated, and it worked right away, and the data made sense. That was when we thought that we were doing something new and we could get the experiment to work. With this we were able to measure the parameters, and even wrote a paper on thermal physics of the measurement. The next step was to build the dilution refrigerator, and then make everything compatible with an ultra-low temperature experiment. And step by step we improved the design and got it to work well. I vividly remember taking data one night where I was monitoring the switching events on the oscilloscope, and you could see three clusters of the switching events, which turned out to be the three quantum transitions of the device. So, it was exciting to see in a clear and persuasive way that this macroscopic system was behaving in a quantum way. We went on to measure all the parameters of the device, and to show both the tunneling rate and these resonances were in very good agreement with the predictions of quantum theory.

Zierler:

In addition to John Clarke, who else was on your thesis committee?

Martinis:

John Clarke of course, but I have forgotten the other professors. That’s a good question.

Zierler:

Was there a theorist?

Martinis:

I do remember that Eugene Commins was on the committee.

Zierler:

Okay.

Martinis:

Commins taught my graduate quantum mechanics course, so I thought he might be interested. Also, he had done some initial experiments to test the Bell inequalities, and he taught about this in his course. So, he was aware in the mid-1980’s of the ideas that would become the field of quantum information. I thought he would enjoy learning about an experiment testing a basic assumption about quantum mechanics.

Zierler:

Now, was Michel Devoret part of your consideration to go to Saclay for your postdoc?

Martinis:

Oh, definitely, as well as Daniel Esteve and Christian Urban. You know, I really enjoyed working with Michel, as he was a great mentor. He went back to Saclay to set up a lab to do this kind of physics, so it was a natural place to go.

Zierler:

What was going on at Saclay? What were some of the exciting stuff that you joined when you got there?

Martinis:

Oh, I just wanted to say one interesting thing. You know, these were young scientists, who were then unknown.

Zierler:

Yeah.

Martinis:

So, the typical postdoc is to go to an established lab to make sure you can get new and exciting results right away, as the next step in your career. So, going to Saclay was risky, especially going to Europe before the internet. But I wanted to work with these really smart experimentalists to continue to explore this new field of quantum devices. And I thought I could learn a lot from them, even if we were still setting up a lab.

So, at the time there was another paper by Anthony Leggett describing the physics for arbitrary dissipation, which could be expressed as what happens if you have time-delayed dissipation. We thought we could use this to measure the time it took to tunneling under the barrier, which was an idea people were still trying to understand at the time. This also represented the next level after my thesis experiment to understand the effects of a complex microwave environment around the quantum device. So, the group designed a nice experiment that had string pulling on a microwave load on a delay line, so that you could directly measure what happened versus the time-delay of the load. It was actually a precursor of the many interesting experiments later showing the effect of resonators on qubit energy levels.

Zierler:

What was exciting at Saclay? What were some of the bigger things going on that you might not have gotten had you stayed in the United States?

Martinis:

Well, it was mostly working with the people there who are really good scientists and doing experiments in this new and exciting field. But one of the interesting things that happened is the discovery of high-Tc superconductors.

Zierler:

Yeah.

Martinis:

As was happening all over the world at the time, we had meetings discussing the latest results and what could be simple experiments to test the superconductors. A scientist very close to Saclay, in Orsay, was making these ceramic oxides, so we were able to get some YBCO material very quickly. So, we decided to look for the Josephson effect, which could tell the charge of the superfluid and thus if it was related to regular superconductivity. So, we did the craziest simple experiment. We took a chunk of this YBCO crystal and screwed an aluminum needle into it, pushing it up against the YBCO. I was thinking that this was never going to work- how are you ever gonna make a tunnel junction? But it turned out that this point-contact was touching a grain in the ceramic, and the grain then made a Josephson junction to the rest of the YBCO. We found a clear Josephson effect, and the charge inferred from the magnitude was in agreement with the standard theory of two electrons. This might have been the earliest experiment to see the Josephson effect. The interesting thing is I was able to go to the March meeting and be part of this very famous session on high-Tc, and I gave a talk at like four in the morning showing this data. This excluded some new theories going around at the time, so it was nice to help people figure out what was going on.

Zierler:

How long were you in France for?

Martinis:

I was there for about a year and a half. After a year I married my wife Jean and she came to live in Paris. But in the end, we wanted to come back to the U.S., to be in our home country. I also wanted to start my own group. So, at the time people were thinking about the physics of single electron tunneling where you could count electrons in small tunnel junctions. And NIST in Boulder, Colorado was interested in it. They decided to hire me to start up an effort to do that. And it was a nice fit because NIST is a measurement laboratory, which is something I enjoyed doing and a really good fit. And I could continue to work on these quantum devices. At NIST, the long-term vision was building an electrical standard based on counting electrons. So, I liked the mission of not just doing good science, but in building a useful instrument. This idea of building something from the physics you’re studying powerfully guides your research and tells you what is important. This also happened with quantum computing, trying to build a quantum computer forces you to understand what basic research needs to be done, what experiments to do.

Zierler:

Did you have an affiliation with CU Boulder, University of Colorado as well?

Martinis:

Not officially, but I was able to have students through CU Boulder. That was very nice, and I liked mentoring graduate students. I also had postdocs who came through NIST, and we were able to hire some of them as NIST staff.

Zierler:

What from your research in Saclay did you bring with you? And what were new opportunities, new research, when you actually got to Boulder?

Martinis:

Well, the new research was based on operating devices with very small capacitance tunnel junctions, so that the charging energy of the tunneling of a single electron controlled the physics. This seemed to be where the field was going, so I changed directions a bit. My first idea on measuring a single tunnel junction did not work out so well, but then the Saclay group figured out how to make an electron pump and turnstile, to count electrons, so I switched direction to that good idea. Then the research really started to take off. We first worked out how to measure and use these devices in a very precise way, in the part per billion level. And we eventually were able to calibrate a capacitor by counting electrons at the parts per million level. During this time Michel Devoret came out to visit, and I worked with him on some other experiments.

Zierler:

What was the budgetary environment like at NIST? Did you have all the resources that you need? Was it tough at times?

Martinis:

I had the resources, and the thing I really liked was I was able to spend most of the time in the lab, doing physics, and not worrying about grant proposals. But I did get some additional grants from both inside and outside of NIST, and I was able to start new projects and try out new ideas. So, I started a project on microcalorimeter detectors to detect x-rays with really good energy resolution, which eventually turned into a large project. I was able to start a project on noise thermometry through some NIST funding. I was also able to start a project at NIST on quantum computing. So, there was a great environment there where you could really do new projects, do basic physics and also build instruments. Plus, I was still in the lab. Unfortunately, because I was good at this, in the end all of my funding was coming from outside, temporary sources. It required getting a lot of funding, and it seemed unstable to me. So, this was a weird situation, but at least I thought my internal funding couldn’t go below zero.

Zierler:

(Laughter)

Martinis:

Right? And then they decided they needed to hire a manager for me who would, of course, start telling me what to do. But then I had to raise the money to help pay for his salary.

Zierler:

Yeah, yeah.

Martinis:

I tried to explain that this was not so good, but that’s what they wanted to do. And that point, I started getting offers at universities based on my quantum computing research. The one from UC Santa Barbara, was really good, and it gave me a chance to work with Andrew Cleland, who was the student after me in John Clarke’s group. And I decided for the amount of money that I’m raising I could have a big university group and do exactly what I wanted. Also, I was over forty and still spinning photoresist in the cleanroom, and I figured it would be fine to stop spending so much time in the lab.

Zierler:

John, a general question, just the sort of the state of the technology at the time of this transition in your career from NIST to Santa Barbara. What were some of the advances both in computers and in instrumentation that made new research possible? And to what extent were you aware of limitations in both computers and instrumentation in terms of ultimately where you wanted to see your research headed at this time?

Martinis:

Yeah. electronics has always been my hobby from high school, and I’ve always built custom electronics and built new instrumentation. Over my career I have designed new electronics that you couldn’t buy that would allow you to do an experiment in a unique way. I have also built custom dilution refrigerators. That is my niche skill set, so all these things are just part of what I have done for good experiments. Another example is the complicated control systems that are used now for qubits. I was making that in the mid-2000s at UCSB using FPGAs. I even taught the digital electronics course using these devices. So, this all started in the mid-eighties with simple electronics, and the complexity just built up over the years.

Zierler:

Do you have a specific memory of the first time you came across the term quantum computing?

Martinis:

Yes (laughter). So, that’s a good story. It happened when I was about ready to graduate from UC Berkeley. There was a conference at UC Santa Barbara on the macroscopic quantum tunneling phenomenon, the subject of my thesis, and we were able to speak about our latest data. So, it was a very exciting conference. And then the last talk of the conference was by Richard Feynman on quantum computing. So, that’s where I first learned about it.

Zierler:

Is your sense that Feynman coined the term?

Martinis:

I don’t know the history of the term, but it was clear he’d been thinking deeply about quantum computing seriously for many years. What was amazing is that the organizers of the conference connected Feynman’s talk to what we were doing with these superconducting qubits, which were basically manipulating and measuring a single qubit, okay? I am not sure that the word qubit had been invented at the time. So, anyway, he gave a talk on quantum computing and it was of course very exciting to listen to Feynman talk about it. He was very excited about it, and you could tell he loved the subject. But I’ll admit that I did not completely understand all that he was talking about, since it was a bit abstract for me at the time.

Zierler:

A-ha.

Martinis:

But I understood it was exciting and a good physics problem. So, after his talk, he was absolutely mobbed by people asking questions because this is obviously an interesting idea. Feynman was in the center and then the senior professors were around him asking questions, and then the junior professors and the postdocs farther and farther away. As a graduate student, I could barely hear what was going on.

Zierler:

Yeah.

Martinis:

Okay. Although I didn’t understand it well at the time, it was clear this was a really interesting field and very important. And that everyone would be interested in it. So, I always had in the back of my mind to work on this subject. But you know, at the time it was still quite theoretical, and I didn’t know what to do, other than to work on quantum devices. So, it took about ten years for the theory to be developed enough so that the government was starting to fund it. And because of my thesis and quantum devices work, they contacted me to see if I was interested. So, I thought about it a bit and came up with an idea on how to isolate the qubit for a longer coherence time. This was a critical idea. I did not understand how to build a quantum computer at the time, but I at least knew how to get started. The first task was to learn how to manipulate a single qubit and do the microwave control, and to measure it properly. And it was just a step by step process to understand what to do.

Zierler:

So, by the time you got to Santa Barbara is that to say that your ideas at this point were sufficiently well-formed? That you could start to think in concrete terms about what it might even look like to build a quantum computer?

Martinis:

Actually, yeah, that’s a really good question because the emphasis did change. So, coming to Santa Barbara we knew how to build qubits and measure them, but we had to make the qubits better. It was a matter of better materials, better control and learning how to put together multiple qubits and resonators. For the individual qubits, materials are hard since for each qubit you are putting together are circuits with billions and billions of atoms. The atoms are connected randomly, and you have to work out the physics for amorphous insulators, which turns out to give a lot of energy loss. It was a matter of finding new materials with lower loss, but in the end, it was more just eliminating these insulators as much as possible. For qubit control, we needed to generate precise waveforms on the nanosecond time scale. This we did by making custom electronics out of FPGAs connected to high-speed digital to analog converters. It was great to collaborate with Andrew Cleland who championed experiments to connect qubits to resonators to make very complex photon states in the resonator. All this was possible using precise control using our electronics and a lot of new calibration software. Many other groups were making progress.

The Yale group with their 3D transmons had long coherence times. But I wanted to stay with our phase qubit, which didn’t have the coherence of the transmon, because I thought it would connect up to other qubits in an easier manner because it had much larger capacitance. The error correction methods at the time, based on the Steane code, had long-range and complex connections, which seemed to need large capacitance qubits. However, all of this changed when a postdoc, Matteo Mariantoni, started talking to Austin Fowler about the surface code. They explain how this error correction did not require super low error qubits, especially since it could be put together in a 2D array with only short nearest-neighbor connections. I got very excited about this idea since it seemed very practical and something that we could actually build soon. And then we didn’t need this large capacitance of the phase qubits, so we could switch to transmon devices which had better coherence. The qubits at the time had about one percent error, reasonably close to the surface code requirement of 0.1%, but much closer than the 0.01% needed for other types.

So, Matteo, Austin, Andrew and I got together to understand this better and write a paper on how it all worked. Now the theory papers at the time were hard to understand, so the idea was to translate all of this work into something that a graduate student could understand. Andrew Cleland did most of the writing since he writes very clearly, and in the end put together a fifty-page Physical Review A paper. My job was mostly to complain that it was not clear enough yet and to add some simple explanations. It’s a very interesting paper because most of it is just a review, you could even say nothing new, but it is heavily cited because it explained how the surface code works so that people in the field could understand it deeply. We are happy that this paper was cited by Physical Review A as one of the top twenty-five papers in the last fifty years in terms of impact. But the important step here is that once I understood the surface code architecture, I knew what we needed to build to make a quantum computer. I knew what to do next. Okay?

Zierler:

And what is the timing of when Google enters the picture for you?

Martinis:

Well, not at first. After this I knew what to build, so I redirected the project. And it was a little bit hard because our funding was to make phase qubits. I argued that everyone at the time was making 3D transmons because they were the best qubits. But I wanted to make 2D transmons on an integrated circuit because they would scale better. So, it was possible to change, and it’s great that the funding agencies let me do that. So first we had to learn how to make transmons and measure them in a new way. At the time, most 2D transmons had a coherence time of less than five or ten microseconds. Using our knowledge of making qubits, we were able to consistently make them with a coherence of twenty to forty microseconds, which was not better than 3D but long enough to do interesting things. So, at once we started working on demonstration experiments towards the surface code. A simpler experiment was to just make a 1D array so you could at least test the bit-flip part of the surface code, and then leave for later the full bit-flip and phase-flip experiment of the 2D array.

So, the first experiment we did was five qubits, which was enough to check if there’s problems with stray interactions or crosstalk. Just to test the basic architecture and the new ideas we had on how to make the gates and do gates on all the qubits at once. And what was amazing about that experiment, it was the first time in my career where the experiment came out better than I thought it would (laughter). There was a lot of work there, but because we had been doing this for ten years, we were able to figure out all the physics quickly. And then in this particular experiment we got gates in a multi-qubit system with fidelities below the surface code threshold, for the first time. And you know, even after ten years it’s still a good result. And then after that we scaled the design to nine qubits to do an actual surface code test. So, during this time when we were getting these really good results, Hartmut Neven contacted me about moving the project to Google.

Okay, there’s a back story to this. Hartmut had started a quantum AI group at Google, first by buying a D-Wave quantum computer. And they were doing experiments to test and understand if it would give them any quantum speedup. One test was to run a randomly connected circuit, comparing the D-Wave quantum computer with numerical simulations, Okay? And initially it looked like the D-Wave was solving it better. This got everyone excited, but then Matthias Troyer, who’s an expert on these simulations, wrote a very powerful simulation that showed that a good computer laptop was as good or better than the D-wave machine. So, it got people at Google excited about the potential, but it’s not yet good enough. During this project I was part of the team, understanding the comparison from an experimental point of view. So, Matthias Troyer introduced me to Hartmut. D-Wave was really building something interesting, finding non-trivial low energy states, and I thought it was good for scientists to look at this scientifically and understand what was happening. And part of it was how do you test whether a quantum computer is working properly or powerfully? And I really liked that Hartmut was trying to understand deeply what this hardware was doing. So, in the end Hartmut decided that maybe Google needs to make its own hardware. And since he knew me from this project, he invited me to come give a talk at Google. The talk I gave was on the five qubit results and what we were building to test nine qubits. And then after the talk, he asked if me and my team want to come to Google to build a quantum computer. So, we strategized about this for a couple of months, and the idea seemed better and better over time, which is what you want. The longest delay was the six months for the lawyers to come to an agreement (laughter). After that we started moving over to Google.

Zierler:

John, what was your sense initially? So, what I have in my mind right now is the model from Bell Labs where it’s industrial research, but there’s absolutely a basic science component to it. To what extent were Google’s motivations at least partly motivated by the basic science? By just supporting this without having any real concrete ideas at this point about how to apply this, how to monetize quantum computing for Google’s bottom line eventually?

Martinis:

Yeah. In fact, that was basically why I talked to Hartmut for three months.

Zierler:

Yeah, yeah.

Martinis:

I wanted to understand what Google wanted to do, what were their short-term and long-term plan. And I had to decide if what we could do with hardware would fit into their plans. It was not just about getting the funding. One of the variables is that we were getting the five qubit and nine qubit device to work and I kind of understood that we could scale up and build a surface code array. So, it looked like this might really work. It’s like I could imagine the Sycamore device and that quantum supremacy would work, even though I did not know the details yet. It was just working. With this in mind, during this time I applied for some renewal government funding. And the reply was that the government didn’t want me to build more than 10 qubits. Now we were already building nine. They wanted me to focus on coherence and infrastructure, which I know how to do, but for me the reason to focus on infrastructure and coherence is to build the quantum computer. That goal tells me what research to do. This all made me consider that maybe I needed to get out of the field because if I can’t build a quantum computer, I don’t know the science and systems engineering you have to solve. I understand that the government might want to build a quantum computer in secret, but I thought that it was too early to clamp down on open research because we still had to figure out many, many things. So, I realized that the model of government funding at that time was not going to work for me.

My thought was that I needed to go out and meet a rich Silicon Valley billionaire, pitch quantum computing, maybe start a center at the university. Or maybe a startup. And then Hartmut called. It’s funny since afterwards I realized that Google was a much better way to fund it (laughter). It was kind of nice that all of this fell into place. I talked to Hartmut about the science we needed to do, and he understood it was a long-term project and that we would start small and see how it would go. He emphasized that working for Google was good, the funding was there, and we could really focus on doing the physics. And that was all true when I moved the project to Google.

Zierler:

What was the negotiation for Santa Barbara? I mean, it’s pretty unique to go down to ten percent from one hundred percent What were some of the negotiations to make this feasible and realistic?

Martinis:

Well, I just told them what we wanted to do and why I had to go to Google. I also explained that the long-term funding from the government was probably going to be a problem.

Zierler:

Yeah, yeah.

Martinis:

And it made sense for Google to do it. The project was going stay in Santa Barbara, we would use its clean room, and we would contribute to the economy of Santa Barbara. There were a lot of good reasons. At that time, I had basically decided in my mind that this was the right step to take and I would have to step down from UC Santa Barbara if I was not able to do it. Also, I was able to keep my students until they graduated. The biggest concern was over IP and separating out UCSB and Google. In the end, everyone was supportive, and it worked out well.

Zierler:

Were there any tensions early on about what Google may have wanted you specifically to do and what you wanted to do as an academic, as a scholar?

Martinis:

You know, I’m a professor. But I’m a very unusual professor, a different kind of academic. Maybe more of an engineering school professor. I’m a definite optimist, but most academics are indefinite optimists. So, they want to do research, but I want to build a quantum computer. This focus is unusual within academia, at least at that time. It’s always been hard for me to fit in to the standard academia model, even though I’ve been successful. So, my goal is to build a quantum computer, not to just write scientific papers. Now it turns out that you have to do basic research and write the papers to build this machine, but papers are not the goal. So, the move to Google was to focus on building a useful quantum computer, that was our mission statement. Now the hard part of such a concrete goal is that you must continue to get results, or you will lose your funding. But then if you don’t get results, maybe it won’t work, and you should change research anyway.

Zierler:

What access to collaborators did you have at Google that you might not have gotten in an academic environment? You know, the stereotype of the genius Googlers that are there. To what extent was that a reality for you?

Martinis:

Well, first of all, at UCSB we had collaborators, but in terms of hardware we were doing it all ourselves. And that’s the basic model we started with at Google. The project started small and we needed to hire physicists, mostly from my team of students and postdocs at UCSB. And of course, Google already had quantum theorists that could help too. With our resources and job security it was possible to hire really good people and focus on just building the quantum computer. We were intensely focusing on building a 2D array of qubits and all the infrastructure needed to get that to work.

Zierler:

When in the story do things start to separate a little bit in terms of your vision and Google’s vision? How do things sort of come to an end?

Martinis:

Yeah, that’s a very hard question. I’m not even sure myself what the heck happened. As background, the quantum supremacy experiment was very stressful to the group. It wasn’t even that popular at first, in part because the experiment was hard and could fail, and we were staking everything on it working. Also, it was natural for all of my students and postdocs to want to be more independent, and that was hard with me leading the group. It’s kind of like how unhealthy it would be for your children to never leave home.

I also think about science differently, like this definite optimism talked about earlier. I am an intense person with a loud voice, especially when I am excited about physics. So, a number of the hardware leaders did not have a good opinion of me as a scientist or leader. Right after we got the quantum supremacy experiment to work, the hardware group insisted that they lead themselves. So Hartmut and higher-level management had me step down as hardware lead. After this I became chief scientist, but it was difficult because decisions went slowly, and I felt I had zero to negative influence in the group.

Scientifically, this was hard since basically no one was talking to me about physics anymore, and people asked me not to come to some sub-group meetings and cut down on outside talks. I was worried about the direction the hardware group was taking, and I was uncomfortable since my employment at Google implied that I thought all was well. This went against my standards of scientific integrity. So, I tried this for a while, but the politics were getting worse over time, so in the end I felt I had no choice but to resign. This was right at the beginning of COVID, and my wife and I were worried about some health problems from stress. I am near retirement age, so we had saved up enough money, and I was grateful for the option. This is all quite sad for me since with better communication, and some counseling and mentoring, these problems could have been addressed early on when they could be solved. But at least I am happy and productive now as a scientist, and I enjoy collaborating with other people and consulting on quantum computing.

Zierler:

To go back to your earlier concern about the problem of hype as it relates to quantum computers. It’s obvious that Google would love nothing more than to go down in the history books as where it all began, right? That quantum computing is a Google breakthrough. Was that part of the issue as well? That sort of intellectually, stylistically, you didn’t see it in those terms?

Martinis:

No. That’s an interesting question. I didn’t think about it that way. One of the interesting things is that when we talked about quantum supremacy at scientific meetings, everyone thought it was five or ten years away. And the nice thing was the Google management telling us, “Hey, it’s a great project! You have an aggressive timeline. We understand that and know it’s hard. We don’t want to create all this technical debt by rushing things in order to meet this timeline.” Technical debt is where you hack things together to get it working, but then have to go back and fix it afterwards. This is well known in the software industry. So, the management was great in the sense that there was the right amount of pressure put on us. Now the group was working really hard, many hours of overtime, and there was a lot of self-pressure. I saw that everyone was being methodical and doing the science properly, understanding measurements step by step. So, when we built the sycamore chip with sixty-four qubits, it worked from the very beginning. And they learned how to calibrate it better and better and then the experiment just worked. So, the team worked well together, we had built up all the infrastructure properly, and didn’t incur too much technical debt in order to get it to work. The team pulled it off! It’s really kind of amazing.

Zierler:

To the extent that you can separate these things out in your mind and in the narrative, to what extent did your collaboration with Google get us closer to true quantum computing? And to what extent did it just generally move things along and its really still TBD (to be determined) in terms of where all of this is headed?

Martinis:

Well, I think moving to Google and being able to focus technically on the project was the key. Staying at a university, the concern was that promotion was based on your scientific papers, and if you spent a year or two getting some technical detail to work better, that’s not so obviously important. And now it’s mostly about solving all the technical problems, okay? I felt like moving to industry was the right move, and that’s what happened at least in the beginning. So, people were able to focus, and you know, we were able to put all the technology together and get it to work.

Zierler:

In the long term, what opportunities do you think this created in terms of you realizing you wanted to stay in the market world? You wanted to stay in the industrial research world? You wanted to continue to explore these ideas not strictly within the boundaries of an academic context?

Martinis:

I’d been thinking about building a quantum computer for a long time, and in the last year at Google I figured out some key ideas on how to scale up to about a million qubits, which is what you need to do error correction on useful problems. And it was very exciting for me, since I finally had a picture in my mind on how to do it, okay? So, I presented it to the team and people started working on the plan and coming up with some new ideas. But what's hard now is after I’ve left Google, I don’t have any resources to work on this anymore. But there’s a lot of things yet to invent, so I can still work by myself. But I’m in this strange situation right now where I finally know what to build, but I can’t do it.

Zierler:

John, for the last part of our talk I’d like to ask a broadly retrospective question and then we’ll end looking to the future. So, where is the inevitability, if I can use that phrase or that word, that would suggest that your academic lineage, your sensibilities, your interests, your talent, to what extent is there an inevitability that all of this would converge on quantum computing? And to what extent is that not necessarily the case that all of these things could’ve led to a very different kind of career for you?

Martinis:

It’s very interesting because basically what is unusual about me as an academic scientist is I think about systems engineering.

Zierler:

Mm-hm. Yeah.

Martinis:

I think about the long-term goals, and what science and technology that must be developed to get there. As a professor, I would often think about other experiments, “Well, I don’t think this is going to work as something useful. You need to do something more practical.” Academia is about exploring everything in a field, which is an important thing to do. But I am thinking about narrowing down my research to build something that works well.

Zierler:

I’m coming back to the way your father taught you how to think when he was a firefighter.

Martinis:

Yeah. You know, I don’t know exactly where it comes from, but let me tell you something interesting. Growing up, my parents did something that was really good for training me as a scientist. Around the dinner table, instead of asking me what we did at school today, they said, “Did you ask any questions today?” And that actually is a profoundly good thing to do.

Zierler:

Yeah. That’s great.

Martinis:

Not just in terms of encouraging your kids to talk, but also you think about the most important part of science, asking questions, right?

Zierler:

Yeah, yeah.

Martinis:

Do you really understand what’s going on? Does it make sense to you? And challenge conventional wisdom. So, I think that my personal approach to science kind of came from that. And then my father was always building things that were different than what other people were doing. And that didn’t bother him. I thought it was great, he did something unique and even oddly creative. So, it was fine in my career to do something that was different than everyone else. I explored a different intellectual space, and it worked out well as long as I explored in the right direction. If not, like for the phase qubit, I eventually changed directions. In that case it maybe took too long to change, but at least I did it for a good reason and understood why.

Zierler:

Well, if we can do some advance historical retrospection it seems pretty obvious that the road to quantum computing is not going to be one where standard thinking is going to be the dominant trait. I think that’s a pretty safe bet to say.

Martinis:

Yeah, exactly! Thanks.

Zierler:

(Laughter).

Martinis:

In scientific talks I always try to remind everyone that building a quantum computer is hard. And maybe everyone’s too optimistic. But when you look at all the constraints from systems engineering, it’s not so obvious that every qubit implementation will work. But you have to try all of them, and it’s not possible at this time to pick the winners and losers. But it’s a lot harder than you think and we need to learn all of the scientific and systems concepts behind building such a complex machine. I think that if you’re going to build a quantum computer- it’s a decade away, you have to stay in startup mode for ten years.

Zierler:

Right. John, for my last question looking to the future, given the fact that this is something that you’ve considered deeply in your recent work. Best case scenario, how does quantum supremacy get us to demonstrate that classical computing simply can’t do things that quantum computing will do?

Martinis:

Yeah, this gets to the point of the quantum supremacy experiment and why I think it’s a pretty important milestone for the field. The way I’ve explained this is that physicists have been talking about the power of a quantum computer for decades. But for engineers who build big data centers, it’s like how is this little physics experiment going to be better than our huge data center? It looks very magical. That’s because it’s hard to understand how quantum mechanics work, especially when you think about its power of manipulating information. The nice thing about the experiment is you run a well-defined algorithm and you simply compare the run time of a quantum computer and the run time of a classical simulator, which is something everyone can understand. You don’t have to talk about complex ideas like entanglement. It’s a definite algorithm and result. It’s not yet useful, but scientists and engineers can understand that since they are used to prototypes and test procedures. Everyone worked really hard to write the paper clearly. My goal when writing the first draft was to make it understandable to Silicon Valley and Google executives, to show them this technology had progressed so that it deserved more serious funding.

At the scientific level, we were testing in a practical manner whether the ideas of quantum mechanics and computing were really correct at a huge scale. Physicists expected it to work, especially the theorists, but the job of experimentalists is to test whether it actually works. And on the practical side, how do you know that there isn’t some experimental noise source or correlation or something you didn’t understand? Especially if a company will put a few billion dollars into making an error-corrected quantum computer. So, for me, the most important part is the test of quantum mechanics. That’s the real practical reason I wanted to do the experiment. In fact, when I gave a talk this year at the APS (American Physical Society) March meeting, I purposely never used the word quantum supremacy.

Zierler:

A-ha, a-ha.

Martinis:

It was all about how to test your quantum computer and know if it was working properly.

Zierler:

There you go.

Martinis:

Because, yeah, that’s the long-term purpose of doing this experiment.

Zierler:

Right.

Martinis:

It was really interesting to write the paper with an A and B story, one about quantum supremacy, and the other about testing a quantum computer. It was kind of fun to balance these two ideas.

Zierler:

It’s just exciting to think about you know, the field really being on the cusp of something major and you know, being privileged to witness these things happening in real time. So, it’s exciting! There’s no doubt about it.

Martinis:

Yes, for me this was a culmination of the research I did for my whole career, and especially what the team did in the past ten years. I had a vision to make a useful quantum computer, and it was a great quest to figure that out to put everything together, to build a powerful quantum computer. But we still have to do something useful, so that’s the next challenge. It was great working with the hardware team to do this, and hopefully I can continue to help with the science.

Zierler:

John, it’s been so fun talking to you. I’m so glad we were able to do this and capture all of your recollections for history. Thank you so much. I really appreciate it.

Martinis:

Thank you. It was a great conversation.