Transcript | Quantum Control Engineering with Q-CTRL

Controlling a quantum computer can mean a lot of things. For Q-CTRL, it includes better error correction and tools to help improve the performance of algorithms by a factor of 9,000 times. It also means having a strong educational environment to help solve the talent shortage facing the industry. Find out more about this quantum control engineering approach in this episode of The Post-Quantum World. I’m your host, Konstantinos Karagiannis. I lead Quantum Computing Services at Protiviti, where we’re helping companies prepare for the benefits and threats of this exploding field. I hope you’ll join each episode as we explore the technology and business impacts of this post-quantum era.

Our guest today is the founder and CEO of Q-CTRL. I’m excited to have him here. He’s also a professor of quantum physics. I’ve been following him for a long time, so it’s great to have him here. So, welcome, Michael Biercuk.

Thanks so much. It’s a real pleasure to be with you.

You got to work with Dave Wineland years ago?

I did.

Nobel Prize winner.

It was before he won the prize. He had long been projected to be a winner at some point. He was passed over, in a sense, in the late ‘90s, and everybody knew it was coming. So, every year, we were hoping and praying that it would come to him. It wasn’t until a couple years after I left, but well deserving.

Yes, I remember when I was talking about quantum in the early days to customers, I used his winning it as a big moment. It’s, like, “Look, they’re paying attention to this. This is important — quantum computing.”

Yes, it was a big statement. In a sense, his award was like lifetime achievement. He had done so much to build this field. One of the more niche technical results that he had achieved earlier in his career, he was passed over, unfortunately, but it was nice that, ultimately, they had this first major prize that was going in the direction of quantum computing for him.

I’ve heard lots of good things coming from Australia lately in general when it comes to quantum computing right now, shifting gears to where you are. We might touch on this later, but that whole Nature thing that was published recently, the UNSW, the high fidelity of qubits. I don’t know if you wanted to chat about that at all today, but it’s an exciting time trying to get these machines. And Q-CTRL’s whole goal, it seems, is to make machines that are more reliable and help other companies get there, along with other offerings. Do you want to tell us about this company? It’s a little hard for people to grasp, probably, when you first hear about it, because you do so much.

We do so much, but we do one thing. That one thing is make quantum technology useful. Our focus is on ensuring that any application of quantum technology, whether it’s in computing, like we were just talking about, or it’s in sensing or in clocks or in a range of other things, they achieve their potential. We do that by focus on something called quantum control engineering.

The inspiration is classical control engineering. I know it’s something that may not be very familiar to most people, but classical control engineering, it’s the discipline that makes everything work. The best way you’ll have seen it recently is if you’ve seen these awesome videos from Boston Dynamics of the walking robots that they hit with a hockey stick, and they right themselves. Those are systems that, obviously, they’re walking bipedal robots, and they fall over if you don’t do something to stabilise them. Boston Dynamics is specialised in control theory, control engineering, to stabilise these robots.

We have a similar approach. We use control theory in quantum systems, and systems that obey the rules of quantum mechanics, to stabilise them against all the sources of noise and error and disturbance that tend to break the technology. This is the Achilles heel of everything quantum, whether in computing or other applications: The hardware is very fragile. Our whole existence is based on overcoming that problem, the fundamental challenge in the field using our specialisation in quantum control.

The goal of that kind of control, when you get right down to the qubit level, do you view it as ultimately avoiding as much error correction as possible, trying to get these to work?

We take a very holistic approach. People will have heard, or may have heard, this phrase before: quantum error correction. It’s worth saying what it is in order to explain what our contribution is in this field. Quantum error correction is a mathematical premise that comes from the earliest days of our field, quantum computing. It says that if you are able to build sufficiently big systems that perform sufficiently well, you can use this special algorithm, called quantum error correction, to identify and then repair any hardware errors — the zeros become ones, in simple language.

Now, that’s pretty amazing. A lot of the work that’s gone on in the field has been on trying to improve the so-called encoding technique, the way you protect the information against perturbation and make it so you can read out errors but not destroy the information. That’s the trick in quantum error correction. That’s the trick mathematically. The trick, practically, is how you make it such that quantum error correction actually gives you a benefit. Right now, it does not. All the different bits and pieces of it have been demonstrated, but every single one of them makes things worse. That’s the problem right now.

Our focus is on working through all the practical challenges — not the math challenges, not the things that you write down pen and paper to say, “This is what you could get if you have an infinitely large quantum computer,” but the practical challenges today that will help make quantum error correction feasible and make it beneficial. That starts at the qubit level. It starts at the way you manipulate the individual quantum bits, because quantum error correction is just an algorithm. It’s an algorithm. If it’s an algorithm, that means it has building blocks, like quantum logic gates. So, we can make better, improve, all of those building blocks, but then the process through which this algorithm corrects errors is just feedback control.

Now, that, again, may not be a very familiar term, but the meaning of it is very familiar. If anybody’s driven on the road and you use cruise control so you can take your foot off the accelerator when you’re on the highway, cruise control is feedback stabilisation. There’s a little device that measures how fast you’re going, compares it to a speed that you set, a target, and then either accelerates or decelerates based on whether you’re going slow or fast. That’s feedback control. Quantum error correction is just feedback control for quantum systems. It’s very specialised. It’s tailored to do some fancy things, but it’s just feedback control. That’s something that we specialise in. There’s a range of ways that Q-CTRL makes quantum error correction a practical reality beyond just the math, but it is a core part of what we are working on as a company.

That’s probably the best I’ve heard a guest explain it on this show — well done. I noticed you’ve been — on that level, when it comes to that more down-to-the-component level — you’ve been working with a lot of different manufacturers. Is this continuing on to the present? Are there releases of processors that we don’t know about that are still in the works closely with you guys? 

There are industry road maps. There are all sorts of engagements that are going on right now that, obviously, neither the companies nor we can talk about. What I can say for sure is that Q-CTRL as an organisation has many partners across the sector. We have customers among many of the major hardware manufacturers and R&D teams at national labs and universities.

We have demonstrated our software tools for quantum control on all of these different platforms. We’ve shown some outrageous results. The most recent results — you’ll get the sneak peek here — are that we were able to show, using our techniques for reducing errors in quantum algorithms, up to 9,000 times improvements in quantum algorithms, 9,000 times over the best competitive approach. So, this is not setting things up to fail and then making it better. It’s pushing as hard as we can with everything that’s in the public domain, and then comparing with our tools. The results are just outrageous. It all comes from the power of quantum control to improve all of the bits and pieces that make quantum algorithms function.

That’s amazing. Were you using any third-party benchmarking to run that through a real-world circuit?

Yes. There’s an organisation called the QED-C, the Quantum Economic Development Consortium, of which Q-CTRL is a member. The QED-C published a set of what are called algorithmic benchmarks. If you come from the high-performance computing world, there are things like the LINPACK benchmarks. It’s similar. These are emerging standards for testing quantum computers. We use those benchmarks. Then, we compared how those benchmarks performed with the default or the competitive approaches and then with our approaches, and those results are quite substantial. They vary, obviously, by algorithms. Some of them are better improvements than others. It depends on the nature of the algorithms, but up to 9,000X.

That is a pretty stellar number. Would you say that this is going to affect any of the published road maps soon? Would you say that maybe they’re not even optimistic enough, that maybe things would be even better than what we’re hearing?

The important thing is, how does 9,000X compare to what we need to achieve? Now, in our sector, our view, in a handwavy way — this is a little bit of back-of-the-envelope calculations — we need about 100,000X improvement over what you see right now to get quantum advantage, when quantum systems perform well enough to potentially perform some useful tasks. Most of the industry road maps have focused on one dimension of that, which is, how big does the quantum system have to be? How many quantum bits? How many qubits? That is the thing that only the hardware manufacturers can press on. So, we are excited for our partners to continue rolling out bigger and bigger systems.

But the other part is the acknowledged but largely unaddressed gap in performance that all of those quantum bits have to get better and better and better. What we are demonstrating is that we can make each one better, and then, when we put them all together and we execute an algorithm, we can make that entire algorithm perform better in a way that is more than the sum of its parts.

This is very important: If you only fix the little individual elements, then you don’t realise the full potential of either quantum control or the hardware, because there are things that emerge only when you run the algorithm. There are errors that creep in only when you’re in the algorithm. So, we take care of both parts: errors at the hardware level, the lowest — the gate level, it’s called — and the errors at the circuit level. What we’re doing is allowing those hardware road maps to realise the full potential of the processors, because qubit numbers alone is not enough to deliver what we need.

Would you say this would impact something like quantum volume or something newer like CLOPS? Would it have a real-world impact on those metrics?

Quantum volume is an interesting one. Quantum volume, and these other metrics, they are proxy measures. The idea is, can I boil down the performance of my processor to some very simple metric that’s sufficiently telling to inform whether it’s a good processor or not a good processor? How does my processor compare to yours, etc.?

A lot of work has gone on in the community and shown that those proxy metrics may not give very much predictive power when it comes to running a real algorithm. So, the newer move to algorithmic benchmarks where you have a suite of algorithms — not just one proxy measure, but also a range of measures that give you a broader view of how well the system performs — that’s where we’ve been focusing our attention.

We’ve looked at things like quantum volume as well, but we find that pursuing the algorithmic benchmarking suite is the most informative approach, and frankly, it is the thing that most closely mimics how we evaluate supercomputers today. We don’t have one measure. We may report something like FLOPS — floating point operations per second — but instead, the people who build and benchmark systems talk about bisectional bandwidth and bytes per FLOP and all these other metrics that don’t typically get talked about, but they’re all core to evaluating processor performance. So, we stick with the algorithmic benchmarking part.

I’m a fan of the super-tech, super-mark benchmarking. I don’t know if you saw that, where they take machines through their real paces.

Yes. Sandia did the same thing. The QED-C did this. They did multiple processors, multiple algorithms. We follow established benchmarking approaches. We just demonstrate these multi-thousand-X improvements in the performance of the hardware against those benchmarks.

For example, taking one of your customers, IBM, are you working, then, on their 1,100-qubit machine with them? Are there any final tweaks that are going into that road map? Obviously, we just had 127, and we’re going into the 400 range. I don’t know if you could talk about anything like that, but is there a part of that development process where Q-CTRL comes in and helps it come to fruition?

We have a different kind of relationship with IBM. We’re a user of their systems and a member of the IBM Quantum Network. We engage with them in joint go-to-market and the like. We are not working on their behind-the-curtain machines. We have other partners with whom we do that kind of work, but you might understand, I also can’t talk about that.

Yes. When it comes to moving out of benchmarking and into the other areas you do, I know one of your first products is Black Opal. Do you want to talk about that? Because that’s a very different piece of the puzzle.

It sounds very different at first, but then, when you take a step back and look at our whole business, it makes a lot of sense strategically. Black Opal is a product where we did a soft launch in November last year. We are going to full retail launch in, approximately, March this year. It is an edtech tool. It is a tool that uses a highly visual interface and well-constructed animations and interactives in order to teach people quantum computing, to go from zero to programming real quantum computers. The target market includes students, software developers, executives, analysts, investors, anybody who wants a strategic advantage in terms of knowledge to take them well beyond the very good but rather superficial YouTube videos that are out there to an actual practical understanding of how these systems work. We sometimes refer to it, it’s like the Duolingo of quantum computing.

One very important thing about it is that it’s made by real professionals. Obviously, that comes on the product side. We have fantastic product engineers making it. That’s why the reviews have been so outstanding in terms of the interface and the graphics, but also, real experts on the content.

A lot of the content is developed by a guy named Chris Ferrie, who wrote a very popular book called Quantum Physics for Babies. He’s fantastic at explaining very complicated ideas, but he also is a professor in quantum control. So, he has the same depth of expertise that all the other technical members of our team have, which means that the content we’re delivering is not like what you find in a lot of the popular, or lay, press, where it tends to be superficial and, often, wrong. We’re able to distill very complicated ideas to simple, digestible, modular bits of material, but always have it deeply insightful from, and driven by, experts in the field.

Does this have any particular operational system that it follows? Does it try to get you to be a Qiskit programmer in the end or any goal like that, or does it stay out of the particular language — how far up the abstraction layer?

Yes, we avoid those frameworks. We find that the frameworks change a lot. Frankly, there is no standard. Qiskit is very popular. But it is not a standard, and Qiskit changes frequently.

So, we took the approach of saying we’re going to teach you from the basic physics up to something called an intermediate representation. The language happens to be called Chasm or Open Chasm. It’s a representation for the individual operations that you would encode, that you would program on a quantum computer. You have a gate. You may say CNOT one, two, or something like that for a CNOT gate between two qubits. That intermediate representation is pretty broadly adopted by most of their players. It’s still something that’s under development, and there will be continued refinements to the intermediate representation. However, that is sufficiently hardware- and framework-agnostic that it can plug in to any.

Now, what we’re adding as we go forward with this platform is a range of vertical modules. We’re going to be adding quantum computing for finance, quantum computing for transport and logistics, quantum computing for biotech and pharma. Still, those are agnostic of the programming framework, because we think it’s important that people learn the fundamentals in a way that is not tied to one particular platform. Now, we’re very happy to help people learn those other platforms as well. There are great resources available. The Qiskit textbook is very good if all you want to do is learn Qiskit, but we want to teach people quantum computing, not just Qiskit.

There have been some attempts to make languages that are supposed to be a little more agnostic too. Silq is one, I guess. Now, there’s Twist, which is supposed to be more aligned with traditional programming. There’ll be a new one by the time this airs, I’m sure.

There are great efforts. There’s no doubt that, in the medium or long term, those will prove very valuable in broadening the range of people who can take advantage of quantum computing. We have a view that the fundamental problems right now are primarily related to the performance of hardware. So, we focus our technical efforts there.

We talked before how Black Opal looks different than these technical products that we offer. Why is that? It’s because we found that there are many more people who come to us and are not sufficiently knowledgeable about the technical details of quantum computing to leverage some of our more technical tools or, frankly, to leverage access to quantum computers. They’re not ready.

In order to expand the market and ensure that as many people as possible are taking advantage of this totally new computational paradigm, we wanted to build something that would facilitate the entry of everybody from software developers and students to investors and executives to facilitate their pathway into the sector. For us, it’s a feeder into our more technical algorithmic-enhancing performance tools, but it’s also a great segue for anybody who, for instance, wants to get a job in the quantum industry and ends up working with the hardware-manufacturing group and wants to use our more technical low-level tools as well.

I’ve heard you say publicly that you don’t think everyone has to have a Ph.D. to contribute in quantum computing, and I totally agree. We’re going to have an army of coders with bachelor’s degrees that could be doing some real good in this space very soon. It is good to see this approach.

We want to facilitate that. We want that to happen. We know it needs to happen to drive the field forward. Black Opal is the best way we do that.

Yes. We have a serious skill shortage. Good luck finding someone who is the right blend.

You talked about common modules. I was curious: You have, let’s say, the one with finance. Would that take the form, then, of explaining the use cases — going into each one of them with examples, portfolio optimisation, things like that?

The core material that’s in the tool now and that’s being released before the retail launch, which is in March, focuses on taking you from the basics of understanding superposition and waves and measurements in quantum computing, and noise, the reality of quantum computing, all the way up through programming algorithms. We teach you about circuits. We teach you what’s happening in the circuits. We teach you algorithmic primitives, and then we teach you, “Here’s how you code in this Chasm sense, Open Chasm intermediate representation.”

The next part that follows is, what are the use cases, and can we teach those use cases in a manner that is similar to what we’ve done with fundamentals of programming quantum computers, of understanding how circuits work, etc.? Indeed, we’re in the discovery process right now for exactly what’s going to be in those vertical modules. We’ve identified some partners with whom we may be bringing them forward, but the objective is to showcase not only how quantum computers work in this field but also, what is the relevance? What are the problems that people care about where quantum computing can potentially give a benefit? We want to give those holistic but tailored viewpoints on those vertical markets.

That’s something I’m looking forward to seeing what you do in that finance space too. You said in March, this is going to be retail ready. Do you want to talk about how people access Black Opal in general? That would be in March that they’d be able to buy a subscription?

Well, you can buy a subscription right now. We had a soft launch around the holidays with some great discounting, and the market uptake was just incredible. The feedback we had was unbelievably positive. We know there are universities that are using this in their courses already. It’s not even in retail launch. It’s in public data, effectively. There are universities that are using it in their courses.

That was all the soft launch. We’re going to the full retail launch, which means all of the core material all the way up through programming quantum computers in this Open Chasm language, that will be released in March, but you can get on right now, have access to the first six of the nine technical skills, and begin accessing. If you purchase early, we’ll give you extra time as well to take advantage of this beta period before we go to full retail launch. Then, after that, after the retail launch, we’ll be working on the enterprise version, which includes a lot of these vertical modules.

Will there be volume discounts, or something for enterprise so they can have all their employees join in?

That’s right. The enterprise version of the tool, instead of an individual signs up on the retail platform and just goes to town, there will be user management. There will be centralised tracking of accomplishment, moderated community access, also, potentially, learning management system integration — all the things that you would expect from an enterprise system. And yes, it is structured around volume. It is structured around an organisation — having a large number of seats available, rather than an individual purchasing a seat for themselves.

I’m going to have a link in the show notes, and obviously, now’s the time. By the time this airs, it will be weeks away.

That’s right. It’s Q-CTRL.com. All you do is look on the product page for Black Opal. You can sign up right now. There’s a free version. If you’re unsure, go in right now, and you can play around with a range of the modules that are accessible, totally free. We hope that you’ll upgrade because you’ll love it so much.

Because you’re teaching with this modular approach and not committing to a language, are there any plans on one day having a Q-CTRL product that lets you do that same thing and have it be translated to different back-end targets, like work with algorithms and then push it out to a machine?

Quite interestingly, the core technical result that I was talking about before, this multi-thousand-X performance boost, that is achieved with a prerelease tool that we’ve built. That tool is called Fire Opal. Fire Opal is designed for quantum computing end users — the people who are developers, write algorithms for quantum computers, or application specialists. Maybe you’re just a user of quantum computing in a data science role, or something like that. The workflow is very simple. You just put in your algorithm, we compile it, and we’ll pass it to the back end that matters, and then we give you the best possible answer out. That answer may be up to 9,000 times better based on what we’ve demonstrated previously relative to competitors.

This is a simplified workflow. There are aspects of that that look like what are called cross compilers. We’re not building a cross compiler, but you, more or less, have to make that functionality in the background. The real technical advance here is what’s called error robust compilation. We have a specialised approach to compiling that reduces errors in the execution of the algorithm. That is what gives some of these enormous benefits.

Anybody who wants it should get in touch. We’re currently in a private alpha. We are interacting with a handful of customers, doing technical demonstrations. If you have algorithms where you’d like to get better performance, then please get in touch. We’d love to talk.

This is literally “You heard it here first.” Fire Opal will be available now. You can get a private alpha going.

Right now, it’s private. Just get in touch, and if you have an algorithm, we’d love to work with you to do a demonstration of how much benefit can be achieved.

I might get in touch myself. We’re always doing things like that. Do you then take it right down to the, let’s say, pulse-control level on a particular machine? Is that’s what’s happening?

All of that is done invisibly to the user. The user needs no knowledge of any of that. The user will say, “Here’s my algorithm. Here’s my specialist version of QAOA,” an approximate algorithm for optimisation, or for a quantum Fourier transform, or for Bernstein-Vazirani — whatever it is. All you do is provide that in a simple intermediate representation that’s widely used in the community. Then, we will compile it. We will add error robustness. We will optimise all the quantum logic gates automatically with our AI agents across the entire quantum computer, something we’ve demonstrated already. Then, we’ll execute with our definitions of the circuit and the quantum logic. We will give you the answer, which will be much better.

That sounds great. I want to play with that. Thanks so much, again. I was excited to have you on here. I’ve been following you for a while. It was great to meet you this way.

It’s my pleasure. Thanks so much. I hope you and your audience who are interested in taking advantage of these tools — either the educational tools or the technical tools — please just get in touch. We’re happy to work with you.

Yes, and again, it will all be in the show notes too. Thanks.

Now, it’s time for Coherence, the quantum executive summary where I take a moment to highlight some of the business impacts we’ve discussed today in case things got too nerdy at times. Let’s recap: Q-CTRL aims to make quantum technology useful via quantum control engineering. Avoiding errors, noise and other interference are key steps.

Error correction has been around since practically the beginning in quantum computing as an idea. Q-CTRL is working on improving all the building blocks used to make error correction a reality. The goal is to make error correction a reliable feedback control system, sort of like cruise-control systems in cars that detect when you’re approaching a car in real time and adjust your speed.

Q-CTRL’s techniques are able to demonstrate up to a 9,000-times improvement in algorithm behavior. This approach is available to users in a tool called Fire Opal. This tool is still in private data, but its boost to compiled-code performance is verified by independent benchmarking.

Q-CTRL feels we still need 100,000-times improvement to attain a quantum advantage in practical applications. Reaching this goal will need a combination of steady improvements to hardware and algorithms, the gate level, and the circuit level. To see or measure these benefits, Q-CTRL naturally emphasises the importance of measuring quantum systems with an algorithmic benchmark suite. Metrics like quantum volume don’t reflect real performance.

The company is also addressing the labor and education shortage in the industry with Black Opal. This is a visual-learning environment to take users from zero to programming quantum computers. Like a Duolingo for quantum computing, with premium educational content and digestible modules, it teaches basic physics up to algorithmic fundamentals shown in Open Chasm as an intermediate representation of coding. This agnostic approach avoids using one language only.

I’m excited to see the quantum for finance modules they’re working on.

We’re in agreement that we need more bachelor’s degree coders in the industry.

That does it for this episode. Thanks to Michael Biercuk for joining to discuss Q-CTRL. Thank you for listening. If you enjoyed this show, please subscribe to Protiviti’s The Post-Quantum World, and leave a review to help others find us. Be sure to follow me on Twitter and Instagram @KonstantHacker. You’ll find links there to what we’re doing in Quantum Computing Services at Protiviti. You can also DM me questions or suggestions for what you’d like to hear on the show. For more information on our quantum services, check out Protiviti.com, or follow Protiviti Tech on Twitter and LinkedIn. Until next time, be kind, and stay quantum curious.