Subscribe
Share
Share
Embed
Welcome to another episode of The New Quantum Era, hosted by Sebastian Hassinger and Kevin Rowney. Today, we have the privilege of speaking with Dr. Robert Schoelkopf, Sterling Professor of Applied Physics at Yale, Director of the Yale Quantum Institute, and CTO and co-founder at Quantum Circuits, Inc. Dr. Schoelkopf is a pioneering figure in the field of quantum computing, particularly known for his contributions to the development of the transmon qubit architecture. In this episode, we delve into the history and future of quantum computing, focusing on the latest advancements in error correction and the innovative dual rail qubit architecture.
Key Highlights:
- Historical Context and Contributions: Dr. Schoelkopf discusses the early days of quantum computing at Yale, including the development of the transmon qubit architecture, which has been foundational for superconducting qubits.
- Introduction to Dual Rail Qubits: Explanation of the dual rail qubit architecture, which promises significant improvements in error detection and correction, potentially reducing the overhead required for fault-tolerant quantum computing.
- Error Correction Strategies: Insights into how the dual rail qubit architecture simplifies the detection and correction of errors, making quantum error correction more efficient and scalable.
- Modular Approach to Quantum Computing: Discussion on the modular design of quantum systems, which allows for easier scaling and maintenance, and the potential for interconnecting quantum modules via microwave photons.
- Future Prospects and Real-World Applications: Dr. Schoelkopf shares his vision for the future of quantum computing, including the commercial deployment of Quantum Circuits, Inc's new quantum devices and the ongoing collaboration between theoretical and experimental approaches to advance the field.
Mentioned in this Episode:
- Yale Quantum Institute
- Quantum Circuits Inc. announces Aqumen Seeker
Join us as we explore these groundbreaking advancements and their implications for the future of quantum computing.
Creators & Guests
Host
What is The New Quantum Era?
Your hosts, Sebastian Hassinger and Kevin Rowney, interview brilliant research scientists, software developers, engineers and others actively exploring the possibilities of our new quantum era. We will cover topics in quantum computing, networking and sensing, focusing on hardware, algorithms and general theory. The show aims for accessibility - neither of us are physicists! - and we'll try to provide context for the terminology and glimpses at the fascinating history of this new field as it evolves in real time.
The New Quantum Era, a podcast by Sebastian Hassinger.
Kevin Rowney:And Kevin Rowney.
Sebastian Hassinger:Welcome back to the podcast. Hey, Kevin. How are you? Hey.
Kevin Rowney:Hey, man. How's it going?
Sebastian Hassinger:Good. I'm excited for our interview today. We're gonna be talking with Rob Schoelkopf from the Yale Quantum Institute. Rob is, quite the pivotal figure in quantum computing.
Kevin Rowney:yeah, absolutely.
Sebastian Hassinger:I mean, my own, career in quantum computing started at IBM Quantum in, 2017. I started working, with them and then formally joined in 2018. Jay Gambetta, of course, the leader of the the R&D efforts there came from Yale. So that was where I first heard Rob's name. And then I suddenly suddenly started Dawn on me that, Chad Rigetti, the founder of Rigetti computing also came from Yale.
Sebastian Hassinger:Jerry Chow, who is leading the hardware R&D efforts at IBM, also from Yale. Alexandre Blais from University of Sherbrooke, also from Yale. Even John Martinis, although he wasn't, at Yale, but he had an affiliation with Michel Devoret who was at Yale, and he was the leader of the effort at Google. So really Rob and Yale's fingerprints are all over superconducting cubits right from the beginning. And that's because the Yale's with was the, the institute that that came up with the transmon architecture, which really unlocked the, the practical implementation of superconducting cubits. So,
Sebastian Hassinger:you know, quite quite the, the contribution already to this field.
Kevin Rowney:Well, this is gonna be great. I can't wait. Let's let's dive in, man.
Sebastian Hassinger:Alright.
Kevin Rowney:Well, hello and welcome back to the new Quantum Era. We're we're really delighted today. We're, too excited to have doctor Robert Schoelkopf as our next guest. He's the Sterling professor of applied physics at Yale and also the director of the Yale Quantum Institute. Professor Schoelkopf is amongst the key foundational contributors to the dawn of the what we call the transmon era of quantum computing.
Kevin Rowney:He is also a leading scholar on the subject of circuit quantum electrodynamics. He and his team today are highly active contributors to the advancement of new discoveries in quantum computing. Welcome, doctor Schoelkopf.
Rob Schoelkopf:Hi. It's nice to be here.
Kevin Rowney:Yeah. We really appreciate your time. Thank you so much. You know, there's, it's such a interesting story arc. Right?
Kevin Rowney:The the dawn of the transmon and and what it's then, led to in terms of new advances in in quantum computing, such a a rich and complex landscape. I can you help, our audience sort of understand the the new, the new advancements that might be happening with respect to, you know, the this this architecture and where it might go next?
Rob Schoelkopf:Sure. I mean, maybe we can also start with some of the the arc of the history. I mean, quantum is pretty new, but I count as an old timer, I think, from the space.
Kevin Rowney:You go back some. Yeah.
Rob Schoelkopf:I started my lab at Yale in 1998 when, all we had was, one, trapped ion qubit. And, you know, the basic question there was, could we make something that was like an integrated circuit that would follow the rules of the quantum world with superposition, entanglement, and all of that? And, you know, we were able to answer those questions very much in the affirmative and, you know, kind of to my surprise by the, sort of, you know, 2008, 2009 or so, things had gotten to the point where we could run-in the university lab, you know, algorithms on a few of these qubits. And in particular, this transmon that you mentioned had gotten to the point in both performance and kind of repeatability stability that it was clear you could start scaling things up. And this is when we sort of saw the dawn of the, you know, industrial push into quantum computing with superconducting, qubits.
Rob Schoelkopf:You know, we had a slightly different, take on things that probably the very first thing you could scale might not be the best thing or the most efficient thing to scale. And, in particular, you know, my research then shifted over to trying to find efficient ways to solve the problem of quantum errors and quantum error correction. And so it's, you know, been a progression, a lot of new ideas around things. And now we have this, new approach called the dual rail qubit, which I've been sort of saying is, like, as the transmon is to making superconducting cubits themselves robust and easy to make and, scalable, the dual rail is kind of that solution for quantum error correction in my mind. It suddenly relaxes a lot of the requirements and gives you an entry point where you can start, really building machines that are beyond the NISC devices that we have today.
Kevin Rowney:Super interesting.
Sebastian Hassinger:That's great.
Rob Schoelkopf:Yeah.
Sebastian Hassinger:And and, Rob, like, before you started working in superconductor or or quantum computing per se, you were involved in in single electron, transistor design and and research. Right? Was there was it was that sort of what brought you over to to quantum computing or how are those related?
Rob Schoelkopf:During my postdoc at Yale in in 94 through 97 or so, I was kinda transitioning. I had been working in superconducting devices, but in a group that did radio astronomy, millimeter devices, but in a group that did radio astronomy, millimeter wave astronomy. And, I had also worked for a couple of years at NASA's Goddard Space Flight Center on cryogenic detectors for astrophysics. So I was very, you know, much in the how sensitive a measurement can we make? How do we make a gadget that can detect something no one has ever detected?
Rob Schoelkopf:And, during my postdoc, we kind of found ways to apply some of these microwave techniques to the domain of nanostructures and, you know, tiny quantum, electronic devices. And so it was very good timing and very fortuitous because that was also right around when Shor found his algorithm, showed that error correction was at least mathematically possible. Right. And everyone started, you know, saying, well, how how does one make a cubit? And at that time, I was like, well, I think we know ways to make qubits.
Rob Schoelkopf:It's gonna be really hard to measure them, but I think I have the tools and I know how to do it. So let's give it a try.
Sebastian Hassinger:I just find that interesting as well because, Yasunobu Nakamura had mentioned, that single electron transistors and the work in that domain had been quite, sort of influential in his works, and he was sort of, you know, 1st Cooper box kind of, circuit. So part of the the prehistory of the superconducting cube. It's a it's an interesting connection. And then Kevin mentioned, circuit QED. That's sort of the coupling of the understanding of the quantum dynamics of of photons and microwave, photons to, to the actual superconducting circuit.
Sebastian Hassinger:Right? That's addresses, as you said, the challenges of a readout per se.
Rob Schoelkopf:Yeah. We use the term circuit QED to cover all the set of ideas that, are basically parallels to cavity quantum electrodynamics. That's the source of the name, or cavity QED, which is when you have single atoms coupling to single photons in the optical domain. And what we've sort of been able to develop and this turned into the kind of quantum computers that, IBM, Google, and and others have out there now is, you know, a way to make the equivalent of an artificial atom out of superconducting circuits and Josephson junctions and the equivalent of optical photons, but in the microwave domain just at a few gigahertz. And that combination then led to ways to not only have quantum bits, but have ways to, as you mentioned, control and measure them and couple them together.
Rob Schoelkopf:Right. And so then you have sort of, all the necessary, ingredients at the physics level to think about making a quantum process.
Kevin Rowney:To actually do compute. Yeah.
Sebastian Hassinger:Yeah. Yeah. Interesting. So you you, you said sort of you're sort of focusing on error correction. Obviously, when when Shor's algorithm, was, unveiled or discovered, you know, the the first reaction was great, but you'll never be able to run that because cubits are so noisy.
Sebastian Hassinger:You're gonna need pair of correction. Shor then follows up the factoring algorithm with the first pair of correction algorithm. Real one two punch. I've I've I've yet to find anything comparable in the field, to what he did. But, that was really, you know, an adaptation of classical error correction to a quantum regime.
Sebastian Hassinger:And then you've got Kitaev sort of bringing, a a quantum native approach with toric codes and then, the successors or the surface code. You mentioned that the cubits that you're building out, QCI, are sort of the the transmon for for error correction. Is it more of a, is is it more of a a classical approach, implemented in a quantum regime, or is it a quantum approach, a native quantum approach?
Rob Schoelkopf:It's a quantum approach. These dual rail qubits that we're now building are, like a small logical qubit, or they're an element which is more complicated than a standard qubit because it has the capability of detecting, the errors natively. That's kind of built into the hardware. And they're also, you know, devices where we're seeing really exciting levels of performance. So, you know, we have coherence times that are in the milliseconds, so longer than the standard transmons.
Rob Schoelkopf:We have state preparation and measurement that's a couple of orders of magnitude better. And coming soon, we'll be publishing some, things on, you know, gate performance that's really quite, remarkable. So I think, you know, what we wanted to do was kind of, in our stealth mode originally at QCI, figure out how to get something really robust that wouldn't require you making millions and millions of cubits to have a fault tolerant machine.
Sebastian Hassinger:Right.
Rob Schoelkopf:You don't wanna scale and get, you know, tiny factors of improvement. You wanna have something where when you implement the error correction on top of these dual rail cubits Right. You're getting gains that are, you know, an order of magnitude with each layer of error correction or something like that. Amazing. Yeah.
Rob Schoelkopf:So and and this is really looking pretty good. I mean, the the way we try to explain this is that our strategy is correct first then scale rather than scaling up and trying to do a brute force approach to conquering the errors at some later stage.
Kevin Rowney:And and if I'm not mistaken, some of these are very recent breakthroughs. I mean, press
Sebastian Hassinger:They are.
Rob Schoelkopf:Yeah. I mean, we started with, implementing the microwave photons as really the information carriers rather than the trans bonds. And there were these various ideas such as cat cubits and other things, which then led to, the dual rail, which is really an idea that only comes in the last 2 or 3 years. Right. But is already being implemented at scale at QCI because we've done all the background work.
Rob Schoelkopf:We had the engineering team in place that knows how to build this novel kind of architecture. So it was really, you know, kind of things coming together quite a bit, after a long period of development in the last few years. And the innovation is really moving very quickly now. You know, we've already sort of gone through 4 different ways of doing the entangling gates, each one improving on the last, just in the last year or so. So,
Kevin Rowney:but it's so so so interesting. Again, it feels like a cutting edge breakthrough just emerging in the press today. But, I mean, I I'm wondering for our audience, perhaps they've this term, you know, a dual rail cubit might might be new to some of them. Is there some
Sebastian Hassinger:It will.
Kevin Rowney:I don't know. Yeah. Easy and convenient way to summarize, that that architecture and how it distinguishes from what what went before.
Rob Schoelkopf:Yes. In in transmons, it's basically the information is stored in the presence or absence of a single excitation. The dual rail is just, slightly more complicated. You have 2 oscillators, 2 places you can place a microwave excitation, and the excitation is either in the left or in the right. But the qubit should always definitely have an excitation present.
Rob Schoelkopf:And when you see that there's no excitation present, you've lost the qubit's information, but at least you know. And it turns out that when you're doing, that there's there's sort of 2 implications of this. One is in the near term, you can run algorithms and say, here's a 1,000,000 shots, but these are the 1,000 that actually work. Interesting. Yeah.
Rob Schoelkopf:Okay. And so you can get, higher fidelity, in sort of short depth circuits, and that's really going beyond the noisy intermediate scale machines we have today where you run a 1,000,000 shots and you're trying to pull the signal, out of the noise, and it's a tiny fraction maybe if you're doing a complicated algorithm.
Sebastian Hassinger:Well, in fact, it's done in it's done statistically in post processing. Giant sample set, and you try to discern which ones after the fact might have been a success or unsuccessful. That's right. So the the error detection is kind of a hack or it's a little bit of a cheat.
Rob Schoelkopf:But if it's a cheat that gives you better performance, then you have a leg up for, doing these other kinds of error mitigation or or other types of things. And, you know, it incorporates I mean, part of the thing we've always been building at QCI is focused on solving the problem of error correction, which means you need to be able, even when you're making a surface code or something like that, you need to be tracking the errors, following them in real time and responding. So your control system and your whole architecture has to know about that and be able to do low latency measurements and the like. And then the second implication of this of this dual rail and the error detection is that you can embed those cubits into, various kinds of error correcting codes. And now you have an advantage.
Rob Schoelkopf:It's sort of simpler to correct these so called erasure errors where the photon disappears than it is the usual kinds of errors. And in our dual rails, you know, the remaining errors are an order of magnitude or low or so below this sort of dominant detectable error.
Kevin Rowney:So interesting. So does this maybe open up the possibility of brand new quantum error detection algorithms, or is it rather that this new architecture is so suitable for many of the extent algorithms already available?
Rob Schoelkopf:Potentially a little bit of both. I think it's a very interesting time in quantum error correction. You know, it it started conceptually a couple of decades ago, but there wasn't really contact between theory and experiment
Rob Schoelkopf:Until just recently. And the idea now that you can evolve your code or design your code to match the hardware, and you can optimize the hardware to, you know, get the best out of certain codes. This kind of co design is common in conventional computing, but, you know, we weren't able to do that in error correction because we hadn't reached this stage yet. So I think we're really at a tipping point for quantum circuits, Inc, and also for, error correction and for the field now where, you know, we'll see things that are are again beyond this NISC era. And, you know, we're starting to see things where you really can implement error correction.
Rob Schoelkopf:And the goal here with by, correcting first and then scaling is that we don't have to scale as much. You can be much more efficient. If you're gaining faster, you won't need this massive overhead of thousands of cubits per bit of compute in the final algorithm. Right.
Sebastian Hassinger:You you mentioned erasure cubits or erasure errors where the the photon disappears. Are there advantages for for detecting and correcting, bit flip or phase flip errors in the in the dual rail qubit as well?
Rob Schoelkopf:Yes. So, the structure of the dual rail means that, the phase errors are limited by the intrinsic phase fluctuations of these microwave cavities that we use to store the photon. And that we can't measure. It's at least milliseconds or longer. We've just placed a bound on it.
Rob Schoelkopf:And then, we also understand that at least when it's idling, the dual rail basically doesn't have bit flips. So we've placed the bound on the bit flip time that's sort of a 10th of a second or something. Wow. And so now you feed that into an error correcting code. You give it that knowledge.
Rob Schoelkopf:Right? And it makes the very tricky problem of locating, finding, and undoing the errors, that much, you know, easier because in in the ordinary approaches, like, it's possible, but it's incredibly complex. Yeah. And that's where this overhead sort of expands on you. And, you know, people talk about, well, I don't know his fault tolerant computing a decade away because we need to get to millions of these elements, and we only have a 1,000 today.
Rob Schoelkopf:And we think it's gonna be faster than that.
Sebastian Hassinger:What's your guess as to sort of the the the range for the ratio of physical to logical cubits for for your architecture?
Rob Schoelkopf:So, you know, if you can really gain this, order of magnitude or more as you increase the distance of the code, you know, you can get to these kind of error rates of 10 to the minus 10 with only, sort of 10 layers of error correction rather than 30 or something which people often talk about. So it's, you know, you're talking about an overhead of several 100 cubits per logical, not 1,000 or 10,000. So it's it can be orders of magnitude, efficiencies is what we're going after. Right. And, you know, I think there's also a very interesting, thing, which is, you know, what does it take to get the quantum advantage?
Rob Schoelkopf:We've been talking about regular machines that have no error correction and are just noisy. And then some far off domain where the computer is basically like a digital computer and errors are completely suppressed.
Sebastian Hassinger:Right.
Rob Schoelkopf:Well, we're never gonna get to that final stage. You're always going to use your quantum computer on the bleeding edge of failure. Because if you can reduce the overhead of the encoding and have just a few more qubits, you're gaining exponentially in the compute power you get. And so you're always going to have something where you're trying to use error detection to suppress the leading errors, or correction to suppress the leading errors, detecting the remaining ones, mitigating the ones you can't deal with, and sort of pulling all those tricks out of the hat to get, you know, the answer that you're looking for.
Kevin Rowney:And then the output will always be a sample set from which you analyze, not an answer.
Rob Schoelkopf:I I think so. I mean, I I think you'd you'd you know, we're we're gonna sort of, you know, be in an area where, you know, now with the machines that we're bringing out at quantum circuits, we can explore this kind of domain of partial error correction and how you combine error detection and error mitigation and, you know, making error aware machines and learning how to program them and how to design algorithms with them. So we're excited to have our users get their hands on this and see what the community can do in terms of creatively coming up with solutions that leverage these new features.
Sebastian Hassinger:So Well,
Kevin Rowney:that's exciting. So what we how soon will that become apparent? I mean, we're gonna see QCI up on on the Amazon Cloud soon.
Rob Schoelkopf:Yeah. So we we also have our own portal, and we have an alpha user program. So, we've had our first users already running on some of our small machines. And we've just, had a press release today announcing the launch of a device we call Acumen Seeker, which is an 8 dual rail qubit that has these unique, functions of error detection and real time control flow. And so, you know, we're, you know, now coming, to the industry with something that's, you know, qualitatively different, we think.
Rob Schoelkopf:And we understand that we can scale this this up, but even with, you know, modest sized machines, you're learning new things because of the new Of course.
Sebastian Hassinger:Well, and you mentioned programming this device. How different is gate and circuit sort of, creation and engineering, from from a developing from a a quantum developer point of view?
Rob Schoelkopf:Right. Right. So this architecture and the dual rail cubits have the usual universal gate set of single and 2 qubit entangling gates. And so you can run any algorithm people have drafted in or or the like. What you can also do though is you can insert features like detect, you know, along all the cubits at this stage in the algorithm and find out which ones have errors.
Rob Schoelkopf:And then you can, you know, respond to that. You can say, okay, the top part of this algorithm actually succeeded. Let me only use the information from that part of the computer. Let me, you know, maybe reset this qubit and see if I get some partial answer that's, still usable. And, you know, you have the ability sort of now to do this, feed forward and so on, which by the way is required for fault tolerant computing when you do things like teleporting in, magic states and the like.
Rob Schoelkopf:I guess we're getting pretty technical here, but That's okay. You know, the the, you know, the idea is that, you know, the quantum machine actually requires a sophisticated classical controller that can respond. And, you know, there's also an interesting possibility that people are starting to talk about, which is combining in real time, some, you know, classical compute and, quantum. You know, you can use the results of one of these measurements. Maybe it's not detecting an error, but it's a measurement of some, part of the algorithm, and then you want to respond in real time.
Rob Schoelkopf:So you can make kind of adaptive algorithms as well, even if we're not worrying about errors. And so So
Kevin Rowney:is this variation quantum eigen solvers or an evolution of that idea?
Rob Schoelkopf:So I guess the the VQE or variational eigen solvers would be one instance, but they're really something where you run, you know, maybe a 1,000 shots at least of the QPU. You look at some average output, and then you have a slow loop that's doing some feedback to try and optimize and find, the actual solution to the chemistry problem. What we're talking about is something where within a single shot, you're actually able to make a measurement and and respond. And so you can have kind of iterative algorithms and and the like. Right?
Kevin Rowney:So then the control system is perhaps inside the the QPU, not not outside the system.
Rob Schoelkopf:It's I mean, our control system sits at room temperature outside the refrigerator, but it has sub microsecond latency in which it can measure and then respond. And so, you know, that's still a 1000 times shorter than the coherence time. So there's a lot you could imagine doing in that time.
Sebastian Hassinger:And you just mentioned, you know, the the sort of the the physical implementation. You you've been addressing sort of the scalability of the error correction approach. There's the whole other, several dimensions of scalability for superconducting, cubits, which is, you know, how much can you fit? How many leads can you fit down into a drill fridge? Those types of so yeah.
Sebastian Hassinger:I mean, are what do you see as sort of the the the physical challenges to scalability that you're that you need to address?
Rob Schoelkopf:Yeah. So I think, indeed the complexity of the wiring in the control system is, you know, a problem we know how to solve, I would say, but, requires investment and some hard engineering. It's, you know, another reason for us taking this approach of efficiency is that, you know, we wanna squeeze more compute out of, you know, when you have a 1,000 wires in each cryostat. The other thing that's exciting that people don't talk about that much in circuit QED or superconducting devices is that you can actually convert quantum information into flying photons that travel on a transmission line or a coaxial cable. And, you know, if you imagine some large machine that can be modular having different blocks that you separately build and test, you can put those all in one cryogenic environment and have these microwave interconnects between them.
Rob Schoelkopf:So, you know, that's a way to kind of help with the scaling. You don't have to have everything crammed onto one chip, for instance.
Kevin Rowney:That's really interesting. It's almost like a a bus of, in computer, technical architecture jargon, a bus of communication between separate Exactly. Computing sub devices. Yeah.
Rob Schoelkopf:In 2007 or so, we published a paper showing sort of a quantum bus and being able to connect to transmons and entangle them over sort of centimeter distances where they didn't actually talk to each other. But they were still on a single chip. There have been since then experiments showing, you know, physically converting to, a cable and then going to a really separate module and being able to, generate entanglement, in that way.
Kevin Rowney:Wow. Really fascinating.
Sebastian Hassinger:Yeah.
Kevin Rowney:Well, and so, for these these new approaches, I I wonder, are there, frontiers yet to cross on the engineering to make, these these chips at higher density and higher scale of manufacture? I mean, you know, many of these issues in the fabrication labs are by by no means an easy challenge to overcome.
Rob Schoelkopf:Yeah. That's that's right. I mean, we take a modular approach sort of from the beginning. So our devices are not a monolithic thing on a single chip. They're actually multi chip modules, where, you know, there are a few elements like a transmon ancilla.
Rob Schoelkopf:It's readout electronics and the like on a single chip. And then, you know, the thing that couples you to the nearby devices is a is another chip. And what we're seeing is that's a really nice approach. If you're trying to make a chip with many, many devices and you have to get your yield to the point where all 1,000 devices on a chip yield, that can be hard. Whereas in this modular approach, you know, if we assemble a system, cool it down, and test it, and there are a few devices that are out of spec, it's possible to sort of pull those individual chips and and replace them.
Rob Schoelkopf:And so, you know, my point of view is the size of the block you should be building as you build a quantum computer is always the largest, quantum unit that you can very reliably build and test. Right? And as we get better and better with the fabrication and and our control, that size of that block will grow, but you always wanna be connecting You you always want something bigger. So you always wanna be connecting up, several of these blocks. And so that's also been a a sort of signature of our research and of this approach at quantum circuits.
Kevin Rowney:That modularity that you've designed into the system allows that kind of growth. Yeah.
Rob Schoelkopf:That's right. Yeah.
Kevin Rowney:Wow. Such a such a a cool new frontier. I mean, you know, there there's so many, bits of hype in this space that people, I think, perhaps, make a bigger deal of these advances than perhaps they are. But this this particular innovation, it does feel like it opens up a whole new angle, right, on what could, what could be a a a realistic route towards large scale in in the near term future. Am I am I putting too much spin on this?
Kevin Rowney:What do you think?
Rob Schoelkopf:No. We think this is really a break through that, will allow us to, you know, scale up more rapidly, and that's gonna take, a little bit of time and some more resources. But, you know, I think, this dual rail architecture feels to me like something that's really worth scaling, at this point. So, you know, we tend to be very conservative, with our claims, but, you know, I'm I'm I'm pretty excited about the, recent results we're we've been seeing, which are, you know, coming at a very fast pace as well.
Sebastian Hassinger:From my perspective, everything you're you're claiming has been backed up with actual experimental results and and numbers you can see in a paper and archive.
Rob Schoelkopf:So Yeah. I I think, it's important to, you know, show show what you've got. And, you know, it's going to be a longish road for getting quantum computing there, and we wanna avoid overhyping and under delivering on, promises and, people's expectations. And so as a field, you know, we should be cautious, I think. And, you know, one of the reasons I decided to get into the commercialization space and found quantum circuits was that, I thought we had things to bring that, other people weren't really considering, and that, you know, I sort of had a responsibility to be, like, you know, contributing to this in in a deliberate way and, you know, try to be a good example in our, academic research.
Rob Schoelkopf:And, you know, that's, I think, shown up in, where our graduates, are. They're leading many of the efforts out there in the field, and they've been successful scientists, on their own. And so, you know, we sort of felt like, you know, that approach was the one we should take with quantum circuits and be careful and deliberate and not overhyped things and and wait till we have, you know, real real proof before going forward.
Kevin Rowney:Yeah. So real real path break breaking work here, but, yeah, a sober minded view of of the future.
Rob Schoelkopf:That's the scientist in me.
Sebastian Hassinger:Refreshing! Thank you. Yeah. Yeah. Well, so much of the, the the challenges in quantum computing are are bridging that gap between, the scientific foundational work that needs to be done and the, as you said, the engineering or technological or or even business challenges, required to to sort of get to implementation of scale and practical advantage.
Sebastian Hassinger:Right. So so if we
Rob Schoelkopf:The thing I get asked very often is, like, is quantum computing now science or engineering? And and my answer is yes.
Kevin Rowney:Yeah. Right.
Rob Schoelkopf:Right? And and, you know, I think it's only scientists who have no idea or experience with technology, transfer and development that would think you do, like, science, science, science, and then all of a sudden you throw it over a wall.
Sebastian Hassinger:Hand it off to the product.
Kevin Rowney:Work it out.
Rob Schoelkopf:And so, you know, that's one reason I've maintained my role at the university is I think there are still innovations that, you know, are best explored in the high risk, high reward, you know, it's okay if it fails. And at the same time, if you're, on the industry side, you can't try to scale something that you don't have an existence proof of. Right? You need to know what it looks like, and you need to be nimble to, you know, incorporate new innovations as they come. And that's that's tricky, and that's gonna be hard for the entire space.
Rob Schoelkopf:But, you know, it's it's something that that one can do if you're careful. Right.
Sebastian Hassinger:And in part, I think also the the confusion and sometimes frustration around the the difficulty in predicting timelines in quantum computing comes from people sort of making category errors about, you know, I mean, you know, you can look at whatever the, portable computing from from the first luggable to the iPhone and go, well, that was a series of engineering advances primarily, and, you know, somewhat predictable in kind of time scales from one step to the other. And with quantum computing, you've got, you know, certain tasks required fundamental. Those high risk, high reward breakthroughs to actually figure out how to solve the problem. And those are very hard to put, error bars around that. You know, how much time will that take?
Sebastian Hassinger:Not sure.
Rob Schoelkopf:Right. Right. And everybody knows about, you know, Moore's law, but, you don't think about all the work that went into material science and device physics Right. Yeah. Along along with that that, you know, allowed us to have the amazing miniaturizations and and advances.
Rob Schoelkopf:Yes. So, you know, I think we're gonna need the equivalent of that in in quantum, a robust research environment and, you know, people pushing on the technology side.
Sebastian Hassinger:You brought up worst law. I've heard people refer to the Shulcov's
Rob Schoelkopf:law,
Sebastian Hassinger:being that the, the time of coherence doubles every, 2 what was it? 2 years, I think, is
Sebastian Hassinger:something that
Rob Schoelkopf:was It was 10 x every 3 years. And we and it and it did that from, you know, starting with Nakamura's first experiment in in 98 up until about 2010 or 2015. And things have gotten hard. It's kind of plateaued. Is that so?
Rob Schoelkopf:So, you know, there's been maybe another factor of 10, in the last decade or so, and people have been really concentrating on, you know, the scaling, but, you know, we've also been working hard on the on the coherence. And in our devices, we currently use these these three-dimensional microwave cavities, which can have, you know, longer lifetimes and and much less noise than, the standard, transmon circuits. So, you know, there's always gonna you're always gonna be able to build a better resonator than you can build a qubit, in my opinion. Right. Right.
Sebastian Hassinger:Right. Right. So looking out a year from now, where where would you like QCI to be where you hope you you are in terms of the development of the of the the technology?
Rob Schoelkopf:Right. Well, I mean, we're very excited to have our machines getting into the hands of users and seeing seeing what they can do with them. I think, you know, in in parallel with exploring that, you know, commercial space, we've got some ambitious plans about demonstrations we can do with error correcting codes in in this next, you know, year or 2. And so, you know, I think the race is on to sort of show an error correcting code that really gives a dramatic improvement. Right?
Rob Schoelkopf:So, we did the first experiments with cat codes in 2016 where we corrected the natural errors. It was just idling, just a memory operation, and we could just barely get some gain. We were overcoming the overhead and, like, 10% extra. And, you know, in the recent experiments with a 100 transmons, they now get a factor of 2 gain. But a factor of 2 for a 100 transmons, that's pretty hard work.
Rob Schoelkopf:And so, you know, I think, you know, if one can show error correction that really has, you know, large gain and you suppress the errors by factor of 10 or more, that's really a, you know, tipping point or I mean, it's really the the proof that, fault tolerant computers will be built. Yeah. Right? It's
Kevin Rowney:a whole whole new game at that point. Yeah.
Rob Schoelkopf:Yeah. And and I think that's, you know, as you said, when Shor first introduced error correction, people wondered would it ever really be practical. I think we know the math works. We know that in principle you can do it, and it kinda works. Can it really be done in practice and efficiently enough that you would benefit by putting error correction into your machines.
Rob Schoelkopf:And that's Yeah. That's the thing that we we think we have a clear path to proving.
Kevin Rowney:God, this is so exciting. We we really appreciate your time. Thank you, doctor Schollfeld.
Rob Schoelkopf:Yeah. Yeah. Thanks a lot. Been it's been great to chat with you guys. A lot of fun.
Kevin Rowney:Wow. That one that was just amazing. An another one of these, great interviews. That was, kinda cool, don't you think? Harry, about these brand new architectures for for QC.
Kevin Rowney:The these dual rail systems, man, it's, quite a quite a breakthrough, it seems like. Yeah. I I guess just press breaking just today on the topic. We'll probably get a link up on the, the post That's right. Of this episode.
Kevin Rowney:But, yeah, it does sketch a whole new future, right, of, of what might be a a brand new framework for, you know, a big scale, way better gain, way better error correction.
Rob Schoelkopf:Right.
Kevin Rowney:Still a long way to go. I I really appreciate also he's got, you know, this inspirational view about where things could go, but also a very sober minded Yeah. Attitude about, you know, just setting expectations. There's still, many innovations left to, make progress against before we're we're at a commercial system.
Sebastian Hassinger:That's right. Yeah. We we we do seem to select our guests based on the, clear eyed view of where the technology is today. And and Rob was no no exception.
Sebastian Hassinger:think there are really significant, you know, what I found really exciting is the idea of entering into a new phase of real co design between a theory of error correction and experimental implementation of at least something that is more aware of the error correction approaches that are going to be, you know, that makes intuitive sense that that's a way to drive higher efficiency and lower ratios of physical cubits to logical cubits. And that of course, you know, lowers the bar in terms of the challenges for making these devices at scale skills that will be commercially, you know, advantageous.
Kevin Rowney:Absolutely. I I I also really appreciate his attitude about how this is, you know, a mix of both, you know, physics theory and, you know, engineering practical problems that must, you know, be in collaboration in in a very constant and very, intense way to make progress against this target. It's not just the scientists on one side of the room and the the implementers on No.
Sebastian Hassinger:At Yeah. Exactly.
Kevin Rowney:Collaboration. Yeah. That's pretty cool.
Sebastian Hassinger:I mean, that's that's what keeps me glued to this space, absolutely, is that it feels like a a rewind to the middle of the last century when we were still trying to find what classical computing might be. And it was a direct collaboration between physicists and engineers for the first time back then, around con information theory and computation. We get to do it all over again with the extra challenge of doing it at the quantum level. And that's just catnip for me. I mean, I I can't stop thinking about it myself.
Sebastian Hassinger:So It it's amazing. Yeah.
Kevin Rowney:Yeah. We we we live in in miraculous times. Yeah.
Sebastian Hassinger:Absolutely. Alright, Kevin. Until next time.
Kevin Rowney:Thanks, man. Good touch. See you on. Okay. That's it for this episode of The New Quantum Era, a podcast by Sebastian Hassinger and Kevin Roney.
Kevin Rowney:Our cool theme music was composed and played by Omar Costa Hamido. Production work is done by our wonderful team over at Podfly. If you are at all like us and enjoy this rich, deep, and interesting topic, please subscribe to our podcast on whichever platform you may stream from. And even consider, if you like what you've heard today, reviewing us on iTunes and or mentioning us on your preferred social media platforms. We're just trying to get the word out on this fascinating topic and would really appreciate your help spreading the word and building community.
Kevin Rowney:Thank you so much for your time.