Podcasts

Podcast with Kirk Bresniker, Chief Architect at Hewlett Packard Labs

25
August
,
2022

My guest today is Kirk Bresniker, Chief Architect at Hewlett Packard Labs. In this extended conversation, Kirk and I talk about when HPE will offer quantum computing services, the time and place for quantum computing in the enterprise IT architecture, and much more.

Listen to additional podcasts here

THE FULL TRANSCRIPT IS BELOW

Yuval:           Hello, Kirk. And thanks for joining me today.

Kirk:             Thanks. It's great to be here.

Yuval:           So who are you and what do you do?

Kirk:             So my name is Kirk Bresniker, and I am the chief architect at Hewlett Packard Labs, part of Hewlett Packard Enterprise. It sounds like a big title, but really, what I get to do is to look over our entire disruptive development portfolio at Hewlett Packard Labs, and the technologies and the technologies that we are bringing to bear, and then I get to have a series of conversations, conversations inside of our business groups, what we think is coming next, just over the brow of the hill, but also conversations with partners, with customers, with policy makers around the world, looking at that intersection of technology and opportunity, thinking not only about the possibilities of technology, but what are those ramifications? What should we be preparing for? How can society make best use of emerging technologies?

Yuval:           Is quantum one of the emerging technologies being asked about?

Kirk:             When customers come to our innovation centers and they see I'm wearing the Hewlett Packard Labs T-shirt, and they say, "Oh, okay. So you're from labs. Then what I really want to know about is ..." and they'll usually ask about two things. They'll ask about AI, and specifically about AI ethics, and one of my side hustles is leading the working group that has drafted our AI ... One of my side hustles is leading the global AI ethics principles teams here at Hewlett Packard Enterprise. So I get to have a little nice conversation about that, but the second one is quantum. Definitely they want to know, and they usually want to know one of two ... They want to know two things.

Kirk:             They want to know when. When is quantum going to be material to my computational landscape, and two, where? Where, in what I need to do, is quantum going to be an advantage, or maybe the other way to look at it is, "When am I going to be surprised because one of my competitors has utilized this technology?" And so that's really an interesting part of the conversations we get to have is to talk about their computational landscape. And because, thank you, Professor Shore, because of your work, we have the cyber security, it becomes part of that conversation. So all of them, it's been really interesting for us to begin to peel back and talk about the potential advantages we all are working towards with quantum computing and the acceleration of specific kinds of workloads.

Kirk:             It also leads into the broader conversation about, "Okay, we are heading into those last best Moore's law scaling transistors." So what is that whole range of accelerators and novel ways to approach problems, some of which we're really anticipating potential for quantum, some are ones where we know it's probably not going to be an advantage, but we still need to understand what's next. What can we do to continue to improve performance, improve efficiency? And most importantly, how to improve sustainability? So we can actually use these technologies and know that we are going to make the world a better place by using them rather than worry that we're using a technology to make a decision, and then we find out the decisions that were left are worse off once we've used the technology than before.

Yuval:           If customers ask you when and where, I would love to know the answers to that as well. So when and where would quantum computing be useful, in your opinion?

Kirk:             So the first question I'll always ask them is, "Do you have any chemists? Do you have any physicists? Do you have people who care about the behavior of physical systems at the quantum level, at the plank distance inside of the molecule, inside of the atomic shell of an atom?" And if the answer is, yes, I'm like, "Well, you should probably be beginning to understand and comprehend these things sooner rather than later." But if there's an enterprise customer, then we'll start to talk about potentially areas like optimization or some aspects of next generation artificial intelligence. And if they are interested and want to understand those, then we'll have that conversation.

Kirk:             We'll certainly also have the conversation about cybersecurity, about, "Do you have information that if someone was to get your encrypted information today, is that information going to be still valuable and material, something that you need to protect in 10 years or 20 years or 40 years?" Trying to ascertain where in their risk profile, the recent announcements we had from the NIST team here in the United States, of the first four candidate algorithms for post quantum cryptography. And adopting a new cryptographic scheme is no small feat, and it has a big price tag. And it's also something that you generally have to do in partnership, because unless you're really concerned about having very safe conversations with yourself, it has to be a conversation that's between multiple people.

Kirk:             So how do you begin to plan out and understand if you're in a regulated business, something like the FIPS, information processing standards, that might be something that you're keenly aware of and might expect to need to take action over the next couple years. 2024 might be when we see some of the FIPS incorporating these NIST recommendations. So I think all those things will end up being part of the conversation for us to have. How much do you want to get involved now? Or would it be better to let some of the technologies settle out, let some of the applications evolve, let people understand?

Kirk:             But it certainly is an an interesting case. And frankly, it's also not just on the teams and partners and customers who want to consume quantum technologies, consume the applications that really can only be possible with quantum advantage once it's feasible. It's also nations and regions who want to get in on the supply side of quantum. They want to be developing and creating quantum technologies, whether it's quantum computation, quantum communications, or quantum sensing, and they see this as a brand new opportunity.

Kirk:             It's a level playing field. There's not one region or country, or certainly not company, who has an advantage right now. And so that means that there's an opportunity for everyone, and who wouldn't want to have been in the lead when we went into semiconductors and into transistorized systems and into integrated circuits? And then you see all the benefits. Certainly, we've all benefited globally from that technology, but it's a different return on investment when you are on the supply side of these technologies, as opposed to just the consumption side. Still incredibly beneficial, raised so many people's economic potential globally, but it's also nice to be on the supply side.

Yuval:           When a new computing architecture or new computing way is introduced, it could find its way into multiple places in the enterprise. It could be in my pocket. It could be on my desktop. It could be at the edge, could be at the data center, the super computing center. Where do you see quantum computing fitting in the enterprise-wide IT infrastructure?

Kirk:             I think for us, it's most likely to come in first as an accelerator, and whether that's an accelerator into a supercomputer, accelerator into an enterprise decision information system, I think that'll depend upon the applications. Am I trying to accelerate a bit of quantum chemistry? Or maybe the better way to say it, "Am I tired of pouring contents from test tubes into beakers and doing experimental chemistry, or would I rather do simulated experimental chemistry?" The same way that we used to smash cars into walls to see what was going to happen, to make sure that they were safe. And now it's like, "Who would smash a perfectly good car in a wall when I could just run an almost nearly infinite number of simulations?"

Kirk:             So I think that's that question about, back to applications, "Are you seeking what Professor Feynman originally set us down the path here on trying to understand the quantum behavior systems and physical systems and materials?" And then I could certainly see wanting to incorporate quantum as an accelerator, whether that is to expose it natively as an, "Here's a simulation of a quantum system," or because I have a very hard problem, and it turns out that some of the quantum algorithms would be an amazing accelerator I might underpin an existing science library with. So I think, again, that's where we might see several avenues for quantum, even in quantum for quantum science's sake. It might be underpinning an existing science library. It might be setting up a simulation, and now, finally, getting some amount of scale on the actual quantum mechanical hardware itself, giving insight into the behavior of quantum systems that we want to engineer.

Kirk:             So I think we might see it there. But again, that's different than saying, "Oh, here's where I want to be able to incorporate quantum because I have a business process. I want to take advantage of the ability of this quantum systems to do things like energy minimization, optimization." Some of that is a little less proven. I know we've had things like quantum annealers for quite some time, and there are certainly other competing non-conventional, still classical physics techniques, that might also give those quantum annealers a run for their money. But that's, I think, another example of bringing these accelerators in and then plugging them into the enterprise, more into the enterprise business backbone, as opposed to the business of science that we might also imagine these systems really helping out.

Yuval:           When you talk about acceleration, would I be correct in assuming that you mean accelerating things that might take days or weeks today to hours or minutes, or are you also talking about, "I have 100 microsecond transaction, and I'll make it 25 microseconds"?

Kirk:             Well, one of my favorite engineering mentors was the analog power design engineer when I started at HP. And he said, "An engineer is somebody who could do something for a nickel that would take any fool a quarter." So there is definitely that shaving off of nanoseconds. Anytime you can change the units or change the prefix, and maybe half the value, then that can be valuable. But I certainly think, for us, it is accelerating existing workloads. Here's a workload, and I just want it to run faster.

Kirk:             We had an amazing experience with DZNE, the German National Center for Neurodegenerative Disease Research. They were doing a genomics application, and it was one of their critical applications, and we were on a call with them, and we were looking at their code, and it's on GitHub. And I open up the code on GitHub. We're having this conversation, and the first line in the comments was, "Because it's impractical to hold a human genome in memory," and I said, "Well, that's interesting because here at the labs, we just created a system, a prototype that holds 80,000 human genomes in memory." So sometimes you do need to reset people's aspirations about what is possible.

Kirk:             Now, in the end, when working with them over a couple months, we took an application, which they assume was terminally optimized. They just could not imagine, and they almost think that they proved themselves it was never going to get any better than this. And yet it was true for conventional hardware approaches. Now, in the matter of, I think we'd spent about 150 days with them, so about five months, and we took that application, and by the time we were done, it was running 100 times faster, 100X fast improvement. It was also using 60% less electricity, which was very material to how they were able to pursue their science. But the most amazing thing about it was it didn't just run now in 13 seconds, as opposed to 20 minutes.

Kirk:             Because it only took about 10 or 13 seconds, the scientists would start the analysis, and they would just watch. And they would watch that little wheel spin around, 10 flips of the hourglass over. And while they're watching that happen, they're already thinking of the next simulation to run. So it moved them into this real time, kept them in the zone of science, as opposed to ... It used to be you'd hit the go button and you'd walk away and you'd do something else for 20 minutes, and you'd come back and maybe you'd just wait till the end of the day, you'd batch up a bunch of them. So being able to change the behavior of the science with a novel hardware, because you're able to get through this breakthrough in acceleration, I think that was something that told me ...

Kirk:             That's the same kind of question, same kind of discussion, same kind of possibility we want to have with the quantum accelerators. Try and understand, "Can I take something that you're already doing and really change it so it changes your behavior?" So I think that's one category. But I think there is also that category where there are things that people just have never ... They say, "Well, I know I could do it. But who's got a million years to run on a supercomputer?" And so there are those classes of applications that we might know. You know the work that's involved, and then you never even consider it as a possibility.

Kirk:             For me, what'll be fascinating with quantum isn't just when we accelerate existing applications, but when we enable and democratize access to this technology so people who won't even know, or perhaps they won't even care, they'll just know this is how science is done. "Oh, of course, I will be using this kind of an algorithm." Or if it's science or business or any kind of decision making that will enable people who say, "Well, yeah, I've got a two trillion node graph, and I wanted to do an analysis on it. Well, of course I can do that. Who couldn't do that? And I'll just type out some code in my Jupyter Notebook, and I'll hit the go button. And I will never be daunted by the scale of a combinatoric explosion." And for me, that's where we'll know we've arrived.

Kirk:             In the same way that people wouldn't bat an eye these days about assuming, "Oh, of course, I'm going to launch a cloud based application that's going to go all the way back to my core data systems and all the way out to my Edge systems, because that's just how we do things." We're cloud natives. We are now all becoming so familiar with using AI and machine learning to do incredible things, like the tools that we're using to communicate with each other this very minute. So I think that's when we know that accelerators have arrived, when their ability to affect problem people really care about is just accepted. It's just the way that we do things.

Kirk:             So I think, for me, it'll be interesting to see both things happen. One, those existing workloads made faster and then made faster enough that it changes people's behavior. And then, two, when you get to that other side of that chasm, when it just becomes the way people do things, and we just all are accepting the common everyday miracle that something like a quantum accelerator could provide.

Yuval:           Let's talk a little bit about quantum computing in HPE. So first, for full disclosure, HPE, through, I think, HPE Pathfinder, is an investor in Classiq. Do you know why the investment was made?

Kirk:             Yeah. So, first, we'll give people a little bit a background into Pathfinder. So Pathfinder is our venture capital arm at Hewlett Packard Enterprise, but it's not just a general venture fund where they're just looking for good ideas and getting the money. The real goal of the Pathfinder team is to find innovators outside of HPE, where they have a great idea. Certainly, the money doesn't hurt, but we also want to provide them is engagement and guidance, and where we see a real affinity between their technologies, what we can bring together, and we can imagine a real successful joint future together. And so that was that's the Pathfinder team.

Kirk:             And I'm sure they have their reasons, but one of the great things about being in Labs is that they come to us all the time, and they say, "Hey, Kirk, we have this company. Tell us what you think." And for me, why I was very encouraging for them to make the investment in Classiq is I think, back to that question of democratization, of access. And I think of myself and I think of semiconductor design. I took semiconductor physics. My last class I had to pass to graduate. I already had the job at HP. All I had to do was get through my semiconductor physics class, and the class was notorious for seniors not quite making it through. So I was a little scared. Did my work, learned what I had to learn, and then, frankly, promptly forgot about it because it turns out that I didn't really need ...

Kirk:             It was good experience, but to actually to produce a design, to brew a chip, what I really needed to know was Verilog, and then I had to learn the tools. I had to learn that EDA environment that would then allow me to design abstractly, and then it would go out. It didn't matter which founder we were going to, inside the company, outside the company, whatever process step, the tool vendors had taken care of creating the synthesis and optimization environment, and that's more than just a clever algorithm. It is also accruing the relationships, the knowledge and relationships of all the possible vendors that you want to take this abstract design engineer and then make that possible for them to gain access to all the benefits and the nuances and the variations so that their design is made it up with exactly the right process.

Kirk:             And then when I saw what you guys were doing with your optimizing compiler and the conversations, my first question is, "Have you had the relationship building conversations? Do you have key knowledge and have you set up the relationships so that regardless if I wanted to go to a superconducting qubit or trapped ion qubit, or any of the other modalities that you had already begun those conversations, so you aren't just optimizing in general for an ideological cue, but you are getting the nitty gritty about the real qubits that we have today and what we'll have in the future?" And so for me, that was a fantastic role for you to take on that was so much like what we saw so critical to making semiconductors continue to grow.

Kirk:             Certainly, we can imagine back in the days when people were still cutting, Rubylith tape and laying out transistors in integrated circuits by hand, they had that low-level intimate knowledge, and they did some amazing designs that really were breathtaking in what they were able to contribute. But that's not scale, right? You can't scale that number of designers with that much infinite knowledge, and it also means that there has to be ... There's so many bottlenecks to that system. And when you can begin to clear those bottlenecks, and you can let high-level engineering teams work more abstractly, you can then also let all the low-level teams really focus exclusively on making their low-level hardware and process better. And then there's you right there in the middle, working out so that everyone can be super efficient by themselves and then, together, from top to bottom, we end up with a system that really scales.

Yuval:           Excellent. So just between you and me, when will HPE offer a quantum computer to its customers?

Kirk:             So it's interesting. So here at Labs, we had teams that were looking at nitrogen vacancies on diamond lattice, yet another qubit modality, about 10 years ago, and that team actually stopped that work voluntarily. They put it on the shelf because, at the time, they didn't see that path forward into practical solutions that our customers really needed. One of the things about Hewlett Packard Labs to go all the way back to 1966, when Bill and Dave asked Barney Oliver, who found Hewlett Packard Labs ...

Kirk:             It hasn't been just a general research arm, in the same way that Pathfinder is not a general venture capital arm. We always wanted to be anchored in our businesses and be that one foot anchored in our current business, one eye towards the horizon of the city. What is coming over? What new technologies can be made available? And also, occasionally, it's about it's about new business as well. Where our new areas where an element of Hewlett Packard Enterprises' understanding and knowledge could be applied and create a new market?

Kirk:             So for us, again, we were doing the qubit work a decade ago, stopped that, that team went on to large-scale integrated photonics and has been doing great work in inter-communications and making the systems that are capable, if we think of the cyber physical system, we need to wrap around that quantum system. I think with work like we've done now with Frontier and the Exascale, that's exactly where we want to go next. But that the question is going to be always customer driven. So when are our customers going to say, "Hey, you know what I really need to do? I have this workload, and here's what I need to be able to accomplish. And we would like to understand together when can we bring quantum technology in for advantage?"

Kirk:             And so it's going to come back to that question of when and where? And that's why we've been participating, really about the last 18 months, trying to understand this. And I think one of the things that also changed over the intervening decade from when we were first working at this was our acquisition, first of SGI, and then of Cray. So adding those two companies to already the high-performance computing expertise we had at Hewlett Packard Enterprise gave us access to a scale of super computing and to customers that we did not have before. So I think that's why it's been really great to lean back into this market and to look ...

Kirk:             We're not going to look to go back and start up our diamond latticework again. What we really want to do is to take advantage of all the knowledge we've gained in integration and industrialization, thinking of things like the wafer-scale in AI accelerator that we integrated at Pittsburgh Supercomputing, working with our partner, Cerebras. Integration of one of the accelerators, in one sense, it's about the physics. "How do I even support this incredibly complex device that's so unlike the rest of the IT infrastructure?" And then once you've done that, then there's that cyber physical control system. How do I get signals in and out if I'm solving something by rotating a qubit, a root two over pie turn to the left, that's going to take some time, depending on the modality, a laser pulse, some microwave energy, all sorts of incredible physical manipulation. So how do we turn digital problems into that analog manipulation?

Kirk:             So I think there's that level, and above that, there's that data infrastructure framework. Here I have something that in a couple qubits can represent petabits even zetabits of information. How do I get that information into and out of these systems? If they are so efficient, how do I actually prepare the systems so that we can have those incredibly efficient computational accelerators always operating at peak efficiency? And that's an infrastructure question.

Kirk:             So that's where we are right now, trying to understand how we bring all the expertise we've gained in incorporating novel accelerators into super computing, and then look for partners, partners like yourself on the software side, partners underneath you on that hardware side, and then begin to pull these things together. And then, in some cases, it's that the customer wants, they just want some qubits. They want a quantum compiler, and they want it in their supercomputing center.

Kirk:             And so I think some of those are some of the first customers that are coming to us and saying, "We want to have this," and it's more because they want to be part of the development of quantum technology side. I think as these continue to mature, and as applications and advantage is demonstrated, then I think we'll more likely see these integrated into more general applications where customers aren't just trying to understand how they participate in quantum design. They want to know how their applications can benefit from quantum technologies. And I think that'll be that second wave. But for us, it's always going to be customer driven. What does the customer want? How can we help them achieve those goals? And especially, when it's something like quantum or any of the other accelerators where it really takes a good bit of engineering for these systems not only to work, but to work well and to advantage.

Yuval:           Excellent. My next to last question is you've been following quantum for a long time. What is new, in your opinion, that you've seen in the last six months that customers should know about?

Kirk:             So I think the NIST conversation and those first post quantum cryptography, I think the interesting thing about quantum and cryptography and cybersecurity is that it has prompted a conversation that probably should have happened anyways, about being resilient and understanding the risks in your cryptographic supply chain. Where are we getting those certificates? How are we making sure that, if we had to, we could switch a vendor out? So I think overall just that hygiene and thoughtfulness, the potential of cryptographic break in public cryptography, I think that's been a very interesting thing. And again, I think it's timely for people now that we're seeing some action from NIST that could find its way into regulatory regimes, that a lot of our customers have to be very careful with. I think that's going to prompt an interesting discussion.

Kirk:             But I think overall, in the quantum, and I got to go with you to the Quantum Technology back in June in Boston, and it was funny, because that was the first time I'd ever been to a pure insider quantum conference. And I wasn't sure what I was going to see. Was it going to be all true believers who it's like, "It's all quantum, and it's just a matter of time," or what it was going to be. But I think it was really interesting to see, one, the mixture of people who wanted to know how to take advantage, even in this midterm. "How can I really start to make these technologies come together?" The number of people who are trying to form a regional quantum technology plan, maybe a national quantum technology plan.

Kirk:             Right after I was with you in Boston, I was at the World Economic Forums, Global Technology Governance Retreat at the Presidio out here in California and San Francisco. And that was one of the things that just kept coming through. It's time to make a plan. It's time to consider these technologies and their ramifications, and just be prepared. Prepare to be prepared. Prepare to be agile. Prepare to have an investigation so that as these technologies mature, that you can begin to plot your course.

Kirk:             It's been interesting, after all the conversations we've been having on quantum, one of the things that people ask me, back to that when question, and you even asked me it, right? And I think one of the interesting things with quantum is we are so used to, from 1970 till now, Moore's law, where we could draw a technology graph and we could label the horizontal access in time, in days, weeks, months, and years. And the vertical access we do is always going to be increasing some increasing exponentially increasing performance metric in that more small cases, the number of transistors per square unit of silicon.

Kirk:             And I think that just has set us up to constantly expect, "Oh, that's how I should understand technology evolution. It's time on the horizontal axis, and it's doubling on the vertical access." And the thing about quantum right now is we're seeing constant improvement. But you know it's the kind of thing where it's going to be punctuated by breakthroughs, and so it's not going to be this nice. Every two years, we're going to get twice as many transistors per unit area. And we have worked across the world over the intervening 50 years to make that true for as long as it has been. And so I think part of the thing for people to consider is think of that horizontal access, not as time, but as maybe as milestones.

Kirk:             Okay, when am I going to see this many qubits of this quality? When am I going to see someone demonstrate an application where they can take advantage of that first 10, that first 100, first 1,000 qubits, and then really, not just showing that it's working, is that the advantage of using it is offsetting any potential switching costs. And so I think that's one thing I would ask people to think about is, "Don't just ask the question of when and expect the answer to come in days, minutes, weeks or hours." Ask the that when question is like, "When can I know I should be taking the next step in developing my quantum engagement strategy? Okay, is it this applications available? Is it that this many qubits are now commercially viable? Is it when this piece of the puzzle gets some really good investigation and research?"

Kirk:             And so it's complicated, and maybe someday we'll have a Moore's law equivalent for quantum. But for right now, it is still in that really interesting, chaotic, messy period where things are getting better, but there's every possibility that they might get really better, really fast, and then you want to have had that conversation already with your risk management team, with your development teams. "How can we be prepared and hedge for a possible quantum future, also hedge for some non-quantum inspired, non-conventional, but still classical physics accelerators?" Or there's that combination. Maybe the key to making today's noisy qubits really become effective is going to be a breakthrough in machine learning that allows them to be dynamically manipulated and error corrected even before they're good enough to do it natively. So some combination of these technologies might also be one of those breakthroughs that we really want to say, "Okay, now it's click. It's trying to take that next step in our plan."

Yuval:           So my very last question is, how can people get in touch with you to learn more about the work that you're doing, that HP Labs is doing?

Kirk:             Labs.hpe.com. You can find all of us there. You can find me on LinkedIn as well. So feel free to reach out.

Yuval:           Excellent. So thank you so much for joining me today.

Kirk:             Thanks. It was great talking with you again.

 

My guest today is Kirk Bresniker, Chief Architect at Hewlett Packard Labs. In this extended conversation, Kirk and I talk about when HPE will offer quantum computing services, the time and place for quantum computing in the enterprise IT architecture, and much more.

Listen to additional podcasts here

THE FULL TRANSCRIPT IS BELOW

Yuval:           Hello, Kirk. And thanks for joining me today.

Kirk:             Thanks. It's great to be here.

Yuval:           So who are you and what do you do?

Kirk:             So my name is Kirk Bresniker, and I am the chief architect at Hewlett Packard Labs, part of Hewlett Packard Enterprise. It sounds like a big title, but really, what I get to do is to look over our entire disruptive development portfolio at Hewlett Packard Labs, and the technologies and the technologies that we are bringing to bear, and then I get to have a series of conversations, conversations inside of our business groups, what we think is coming next, just over the brow of the hill, but also conversations with partners, with customers, with policy makers around the world, looking at that intersection of technology and opportunity, thinking not only about the possibilities of technology, but what are those ramifications? What should we be preparing for? How can society make best use of emerging technologies?

Yuval:           Is quantum one of the emerging technologies being asked about?

Kirk:             When customers come to our innovation centers and they see I'm wearing the Hewlett Packard Labs T-shirt, and they say, "Oh, okay. So you're from labs. Then what I really want to know about is ..." and they'll usually ask about two things. They'll ask about AI, and specifically about AI ethics, and one of my side hustles is leading the working group that has drafted our AI ... One of my side hustles is leading the global AI ethics principles teams here at Hewlett Packard Enterprise. So I get to have a little nice conversation about that, but the second one is quantum. Definitely they want to know, and they usually want to know one of two ... They want to know two things.

Kirk:             They want to know when. When is quantum going to be material to my computational landscape, and two, where? Where, in what I need to do, is quantum going to be an advantage, or maybe the other way to look at it is, "When am I going to be surprised because one of my competitors has utilized this technology?" And so that's really an interesting part of the conversations we get to have is to talk about their computational landscape. And because, thank you, Professor Shore, because of your work, we have the cyber security, it becomes part of that conversation. So all of them, it's been really interesting for us to begin to peel back and talk about the potential advantages we all are working towards with quantum computing and the acceleration of specific kinds of workloads.

Kirk:             It also leads into the broader conversation about, "Okay, we are heading into those last best Moore's law scaling transistors." So what is that whole range of accelerators and novel ways to approach problems, some of which we're really anticipating potential for quantum, some are ones where we know it's probably not going to be an advantage, but we still need to understand what's next. What can we do to continue to improve performance, improve efficiency? And most importantly, how to improve sustainability? So we can actually use these technologies and know that we are going to make the world a better place by using them rather than worry that we're using a technology to make a decision, and then we find out the decisions that were left are worse off once we've used the technology than before.

Yuval:           If customers ask you when and where, I would love to know the answers to that as well. So when and where would quantum computing be useful, in your opinion?

Kirk:             So the first question I'll always ask them is, "Do you have any chemists? Do you have any physicists? Do you have people who care about the behavior of physical systems at the quantum level, at the plank distance inside of the molecule, inside of the atomic shell of an atom?" And if the answer is, yes, I'm like, "Well, you should probably be beginning to understand and comprehend these things sooner rather than later." But if there's an enterprise customer, then we'll start to talk about potentially areas like optimization or some aspects of next generation artificial intelligence. And if they are interested and want to understand those, then we'll have that conversation.

Kirk:             We'll certainly also have the conversation about cybersecurity, about, "Do you have information that if someone was to get your encrypted information today, is that information going to be still valuable and material, something that you need to protect in 10 years or 20 years or 40 years?" Trying to ascertain where in their risk profile, the recent announcements we had from the NIST team here in the United States, of the first four candidate algorithms for post quantum cryptography. And adopting a new cryptographic scheme is no small feat, and it has a big price tag. And it's also something that you generally have to do in partnership, because unless you're really concerned about having very safe conversations with yourself, it has to be a conversation that's between multiple people.

Kirk:             So how do you begin to plan out and understand if you're in a regulated business, something like the FIPS, information processing standards, that might be something that you're keenly aware of and might expect to need to take action over the next couple years. 2024 might be when we see some of the FIPS incorporating these NIST recommendations. So I think all those things will end up being part of the conversation for us to have. How much do you want to get involved now? Or would it be better to let some of the technologies settle out, let some of the applications evolve, let people understand?

Kirk:             But it certainly is an an interesting case. And frankly, it's also not just on the teams and partners and customers who want to consume quantum technologies, consume the applications that really can only be possible with quantum advantage once it's feasible. It's also nations and regions who want to get in on the supply side of quantum. They want to be developing and creating quantum technologies, whether it's quantum computation, quantum communications, or quantum sensing, and they see this as a brand new opportunity.

Kirk:             It's a level playing field. There's not one region or country, or certainly not company, who has an advantage right now. And so that means that there's an opportunity for everyone, and who wouldn't want to have been in the lead when we went into semiconductors and into transistorized systems and into integrated circuits? And then you see all the benefits. Certainly, we've all benefited globally from that technology, but it's a different return on investment when you are on the supply side of these technologies, as opposed to just the consumption side. Still incredibly beneficial, raised so many people's economic potential globally, but it's also nice to be on the supply side.

Yuval:           When a new computing architecture or new computing way is introduced, it could find its way into multiple places in the enterprise. It could be in my pocket. It could be on my desktop. It could be at the edge, could be at the data center, the super computing center. Where do you see quantum computing fitting in the enterprise-wide IT infrastructure?

Kirk:             I think for us, it's most likely to come in first as an accelerator, and whether that's an accelerator into a supercomputer, accelerator into an enterprise decision information system, I think that'll depend upon the applications. Am I trying to accelerate a bit of quantum chemistry? Or maybe the better way to say it, "Am I tired of pouring contents from test tubes into beakers and doing experimental chemistry, or would I rather do simulated experimental chemistry?" The same way that we used to smash cars into walls to see what was going to happen, to make sure that they were safe. And now it's like, "Who would smash a perfectly good car in a wall when I could just run an almost nearly infinite number of simulations?"

Kirk:             So I think that's that question about, back to applications, "Are you seeking what Professor Feynman originally set us down the path here on trying to understand the quantum behavior systems and physical systems and materials?" And then I could certainly see wanting to incorporate quantum as an accelerator, whether that is to expose it natively as an, "Here's a simulation of a quantum system," or because I have a very hard problem, and it turns out that some of the quantum algorithms would be an amazing accelerator I might underpin an existing science library with. So I think, again, that's where we might see several avenues for quantum, even in quantum for quantum science's sake. It might be underpinning an existing science library. It might be setting up a simulation, and now, finally, getting some amount of scale on the actual quantum mechanical hardware itself, giving insight into the behavior of quantum systems that we want to engineer.

Kirk:             So I think we might see it there. But again, that's different than saying, "Oh, here's where I want to be able to incorporate quantum because I have a business process. I want to take advantage of the ability of this quantum systems to do things like energy minimization, optimization." Some of that is a little less proven. I know we've had things like quantum annealers for quite some time, and there are certainly other competing non-conventional, still classical physics techniques, that might also give those quantum annealers a run for their money. But that's, I think, another example of bringing these accelerators in and then plugging them into the enterprise, more into the enterprise business backbone, as opposed to the business of science that we might also imagine these systems really helping out.

Yuval:           When you talk about acceleration, would I be correct in assuming that you mean accelerating things that might take days or weeks today to hours or minutes, or are you also talking about, "I have 100 microsecond transaction, and I'll make it 25 microseconds"?

Kirk:             Well, one of my favorite engineering mentors was the analog power design engineer when I started at HP. And he said, "An engineer is somebody who could do something for a nickel that would take any fool a quarter." So there is definitely that shaving off of nanoseconds. Anytime you can change the units or change the prefix, and maybe half the value, then that can be valuable. But I certainly think, for us, it is accelerating existing workloads. Here's a workload, and I just want it to run faster.

Kirk:             We had an amazing experience with DZNE, the German National Center for Neurodegenerative Disease Research. They were doing a genomics application, and it was one of their critical applications, and we were on a call with them, and we were looking at their code, and it's on GitHub. And I open up the code on GitHub. We're having this conversation, and the first line in the comments was, "Because it's impractical to hold a human genome in memory," and I said, "Well, that's interesting because here at the labs, we just created a system, a prototype that holds 80,000 human genomes in memory." So sometimes you do need to reset people's aspirations about what is possible.

Kirk:             Now, in the end, when working with them over a couple months, we took an application, which they assume was terminally optimized. They just could not imagine, and they almost think that they proved themselves it was never going to get any better than this. And yet it was true for conventional hardware approaches. Now, in the matter of, I think we'd spent about 150 days with them, so about five months, and we took that application, and by the time we were done, it was running 100 times faster, 100X fast improvement. It was also using 60% less electricity, which was very material to how they were able to pursue their science. But the most amazing thing about it was it didn't just run now in 13 seconds, as opposed to 20 minutes.

Kirk:             Because it only took about 10 or 13 seconds, the scientists would start the analysis, and they would just watch. And they would watch that little wheel spin around, 10 flips of the hourglass over. And while they're watching that happen, they're already thinking of the next simulation to run. So it moved them into this real time, kept them in the zone of science, as opposed to ... It used to be you'd hit the go button and you'd walk away and you'd do something else for 20 minutes, and you'd come back and maybe you'd just wait till the end of the day, you'd batch up a bunch of them. So being able to change the behavior of the science with a novel hardware, because you're able to get through this breakthrough in acceleration, I think that was something that told me ...

Kirk:             That's the same kind of question, same kind of discussion, same kind of possibility we want to have with the quantum accelerators. Try and understand, "Can I take something that you're already doing and really change it so it changes your behavior?" So I think that's one category. But I think there is also that category where there are things that people just have never ... They say, "Well, I know I could do it. But who's got a million years to run on a supercomputer?" And so there are those classes of applications that we might know. You know the work that's involved, and then you never even consider it as a possibility.

Kirk:             For me, what'll be fascinating with quantum isn't just when we accelerate existing applications, but when we enable and democratize access to this technology so people who won't even know, or perhaps they won't even care, they'll just know this is how science is done. "Oh, of course, I will be using this kind of an algorithm." Or if it's science or business or any kind of decision making that will enable people who say, "Well, yeah, I've got a two trillion node graph, and I wanted to do an analysis on it. Well, of course I can do that. Who couldn't do that? And I'll just type out some code in my Jupyter Notebook, and I'll hit the go button. And I will never be daunted by the scale of a combinatoric explosion." And for me, that's where we'll know we've arrived.

Kirk:             In the same way that people wouldn't bat an eye these days about assuming, "Oh, of course, I'm going to launch a cloud based application that's going to go all the way back to my core data systems and all the way out to my Edge systems, because that's just how we do things." We're cloud natives. We are now all becoming so familiar with using AI and machine learning to do incredible things, like the tools that we're using to communicate with each other this very minute. So I think that's when we know that accelerators have arrived, when their ability to affect problem people really care about is just accepted. It's just the way that we do things.

Kirk:             So I think, for me, it'll be interesting to see both things happen. One, those existing workloads made faster and then made faster enough that it changes people's behavior. And then, two, when you get to that other side of that chasm, when it just becomes the way people do things, and we just all are accepting the common everyday miracle that something like a quantum accelerator could provide.

Yuval:           Let's talk a little bit about quantum computing in HPE. So first, for full disclosure, HPE, through, I think, HPE Pathfinder, is an investor in Classiq. Do you know why the investment was made?

Kirk:             Yeah. So, first, we'll give people a little bit a background into Pathfinder. So Pathfinder is our venture capital arm at Hewlett Packard Enterprise, but it's not just a general venture fund where they're just looking for good ideas and getting the money. The real goal of the Pathfinder team is to find innovators outside of HPE, where they have a great idea. Certainly, the money doesn't hurt, but we also want to provide them is engagement and guidance, and where we see a real affinity between their technologies, what we can bring together, and we can imagine a real successful joint future together. And so that was that's the Pathfinder team.

Kirk:             And I'm sure they have their reasons, but one of the great things about being in Labs is that they come to us all the time, and they say, "Hey, Kirk, we have this company. Tell us what you think." And for me, why I was very encouraging for them to make the investment in Classiq is I think, back to that question of democratization, of access. And I think of myself and I think of semiconductor design. I took semiconductor physics. My last class I had to pass to graduate. I already had the job at HP. All I had to do was get through my semiconductor physics class, and the class was notorious for seniors not quite making it through. So I was a little scared. Did my work, learned what I had to learn, and then, frankly, promptly forgot about it because it turns out that I didn't really need ...

Kirk:             It was good experience, but to actually to produce a design, to brew a chip, what I really needed to know was Verilog, and then I had to learn the tools. I had to learn that EDA environment that would then allow me to design abstractly, and then it would go out. It didn't matter which founder we were going to, inside the company, outside the company, whatever process step, the tool vendors had taken care of creating the synthesis and optimization environment, and that's more than just a clever algorithm. It is also accruing the relationships, the knowledge and relationships of all the possible vendors that you want to take this abstract design engineer and then make that possible for them to gain access to all the benefits and the nuances and the variations so that their design is made it up with exactly the right process.

Kirk:             And then when I saw what you guys were doing with your optimizing compiler and the conversations, my first question is, "Have you had the relationship building conversations? Do you have key knowledge and have you set up the relationships so that regardless if I wanted to go to a superconducting qubit or trapped ion qubit, or any of the other modalities that you had already begun those conversations, so you aren't just optimizing in general for an ideological cue, but you are getting the nitty gritty about the real qubits that we have today and what we'll have in the future?" And so for me, that was a fantastic role for you to take on that was so much like what we saw so critical to making semiconductors continue to grow.

Kirk:             Certainly, we can imagine back in the days when people were still cutting, Rubylith tape and laying out transistors in integrated circuits by hand, they had that low-level intimate knowledge, and they did some amazing designs that really were breathtaking in what they were able to contribute. But that's not scale, right? You can't scale that number of designers with that much infinite knowledge, and it also means that there has to be ... There's so many bottlenecks to that system. And when you can begin to clear those bottlenecks, and you can let high-level engineering teams work more abstractly, you can then also let all the low-level teams really focus exclusively on making their low-level hardware and process better. And then there's you right there in the middle, working out so that everyone can be super efficient by themselves and then, together, from top to bottom, we end up with a system that really scales.

Yuval:           Excellent. So just between you and me, when will HPE offer a quantum computer to its customers?

Kirk:             So it's interesting. So here at Labs, we had teams that were looking at nitrogen vacancies on diamond lattice, yet another qubit modality, about 10 years ago, and that team actually stopped that work voluntarily. They put it on the shelf because, at the time, they didn't see that path forward into practical solutions that our customers really needed. One of the things about Hewlett Packard Labs to go all the way back to 1966, when Bill and Dave asked Barney Oliver, who found Hewlett Packard Labs ...

Kirk:             It hasn't been just a general research arm, in the same way that Pathfinder is not a general venture capital arm. We always wanted to be anchored in our businesses and be that one foot anchored in our current business, one eye towards the horizon of the city. What is coming over? What new technologies can be made available? And also, occasionally, it's about it's about new business as well. Where our new areas where an element of Hewlett Packard Enterprises' understanding and knowledge could be applied and create a new market?

Kirk:             So for us, again, we were doing the qubit work a decade ago, stopped that, that team went on to large-scale integrated photonics and has been doing great work in inter-communications and making the systems that are capable, if we think of the cyber physical system, we need to wrap around that quantum system. I think with work like we've done now with Frontier and the Exascale, that's exactly where we want to go next. But that the question is going to be always customer driven. So when are our customers going to say, "Hey, you know what I really need to do? I have this workload, and here's what I need to be able to accomplish. And we would like to understand together when can we bring quantum technology in for advantage?"

Kirk:             And so it's going to come back to that question of when and where? And that's why we've been participating, really about the last 18 months, trying to understand this. And I think one of the things that also changed over the intervening decade from when we were first working at this was our acquisition, first of SGI, and then of Cray. So adding those two companies to already the high-performance computing expertise we had at Hewlett Packard Enterprise gave us access to a scale of super computing and to customers that we did not have before. So I think that's why it's been really great to lean back into this market and to look ...

Kirk:             We're not going to look to go back and start up our diamond latticework again. What we really want to do is to take advantage of all the knowledge we've gained in integration and industrialization, thinking of things like the wafer-scale in AI accelerator that we integrated at Pittsburgh Supercomputing, working with our partner, Cerebras. Integration of one of the accelerators, in one sense, it's about the physics. "How do I even support this incredibly complex device that's so unlike the rest of the IT infrastructure?" And then once you've done that, then there's that cyber physical control system. How do I get signals in and out if I'm solving something by rotating a qubit, a root two over pie turn to the left, that's going to take some time, depending on the modality, a laser pulse, some microwave energy, all sorts of incredible physical manipulation. So how do we turn digital problems into that analog manipulation?

Kirk:             So I think there's that level, and above that, there's that data infrastructure framework. Here I have something that in a couple qubits can represent petabits even zetabits of information. How do I get that information into and out of these systems? If they are so efficient, how do I actually prepare the systems so that we can have those incredibly efficient computational accelerators always operating at peak efficiency? And that's an infrastructure question.

Kirk:             So that's where we are right now, trying to understand how we bring all the expertise we've gained in incorporating novel accelerators into super computing, and then look for partners, partners like yourself on the software side, partners underneath you on that hardware side, and then begin to pull these things together. And then, in some cases, it's that the customer wants, they just want some qubits. They want a quantum compiler, and they want it in their supercomputing center.

Kirk:             And so I think some of those are some of the first customers that are coming to us and saying, "We want to have this," and it's more because they want to be part of the development of quantum technology side. I think as these continue to mature, and as applications and advantage is demonstrated, then I think we'll more likely see these integrated into more general applications where customers aren't just trying to understand how they participate in quantum design. They want to know how their applications can benefit from quantum technologies. And I think that'll be that second wave. But for us, it's always going to be customer driven. What does the customer want? How can we help them achieve those goals? And especially, when it's something like quantum or any of the other accelerators where it really takes a good bit of engineering for these systems not only to work, but to work well and to advantage.

Yuval:           Excellent. My next to last question is you've been following quantum for a long time. What is new, in your opinion, that you've seen in the last six months that customers should know about?

Kirk:             So I think the NIST conversation and those first post quantum cryptography, I think the interesting thing about quantum and cryptography and cybersecurity is that it has prompted a conversation that probably should have happened anyways, about being resilient and understanding the risks in your cryptographic supply chain. Where are we getting those certificates? How are we making sure that, if we had to, we could switch a vendor out? So I think overall just that hygiene and thoughtfulness, the potential of cryptographic break in public cryptography, I think that's been a very interesting thing. And again, I think it's timely for people now that we're seeing some action from NIST that could find its way into regulatory regimes, that a lot of our customers have to be very careful with. I think that's going to prompt an interesting discussion.

Kirk:             But I think overall, in the quantum, and I got to go with you to the Quantum Technology back in June in Boston, and it was funny, because that was the first time I'd ever been to a pure insider quantum conference. And I wasn't sure what I was going to see. Was it going to be all true believers who it's like, "It's all quantum, and it's just a matter of time," or what it was going to be. But I think it was really interesting to see, one, the mixture of people who wanted to know how to take advantage, even in this midterm. "How can I really start to make these technologies come together?" The number of people who are trying to form a regional quantum technology plan, maybe a national quantum technology plan.

Kirk:             Right after I was with you in Boston, I was at the World Economic Forums, Global Technology Governance Retreat at the Presidio out here in California and San Francisco. And that was one of the things that just kept coming through. It's time to make a plan. It's time to consider these technologies and their ramifications, and just be prepared. Prepare to be prepared. Prepare to be agile. Prepare to have an investigation so that as these technologies mature, that you can begin to plot your course.

Kirk:             It's been interesting, after all the conversations we've been having on quantum, one of the things that people ask me, back to that when question, and you even asked me it, right? And I think one of the interesting things with quantum is we are so used to, from 1970 till now, Moore's law, where we could draw a technology graph and we could label the horizontal access in time, in days, weeks, months, and years. And the vertical access we do is always going to be increasing some increasing exponentially increasing performance metric in that more small cases, the number of transistors per square unit of silicon.

Kirk:             And I think that just has set us up to constantly expect, "Oh, that's how I should understand technology evolution. It's time on the horizontal axis, and it's doubling on the vertical access." And the thing about quantum right now is we're seeing constant improvement. But you know it's the kind of thing where it's going to be punctuated by breakthroughs, and so it's not going to be this nice. Every two years, we're going to get twice as many transistors per unit area. And we have worked across the world over the intervening 50 years to make that true for as long as it has been. And so I think part of the thing for people to consider is think of that horizontal access, not as time, but as maybe as milestones.

Kirk:             Okay, when am I going to see this many qubits of this quality? When am I going to see someone demonstrate an application where they can take advantage of that first 10, that first 100, first 1,000 qubits, and then really, not just showing that it's working, is that the advantage of using it is offsetting any potential switching costs. And so I think that's one thing I would ask people to think about is, "Don't just ask the question of when and expect the answer to come in days, minutes, weeks or hours." Ask the that when question is like, "When can I know I should be taking the next step in developing my quantum engagement strategy? Okay, is it this applications available? Is it that this many qubits are now commercially viable? Is it when this piece of the puzzle gets some really good investigation and research?"

Kirk:             And so it's complicated, and maybe someday we'll have a Moore's law equivalent for quantum. But for right now, it is still in that really interesting, chaotic, messy period where things are getting better, but there's every possibility that they might get really better, really fast, and then you want to have had that conversation already with your risk management team, with your development teams. "How can we be prepared and hedge for a possible quantum future, also hedge for some non-quantum inspired, non-conventional, but still classical physics accelerators?" Or there's that combination. Maybe the key to making today's noisy qubits really become effective is going to be a breakthrough in machine learning that allows them to be dynamically manipulated and error corrected even before they're good enough to do it natively. So some combination of these technologies might also be one of those breakthroughs that we really want to say, "Okay, now it's click. It's trying to take that next step in our plan."

Yuval:           So my very last question is, how can people get in touch with you to learn more about the work that you're doing, that HP Labs is doing?

Kirk:             Labs.hpe.com. You can find all of us there. You can find me on LinkedIn as well. So feel free to reach out.

Yuval:           Excellent. So thank you so much for joining me today.

Kirk:             Thanks. It was great talking with you again.

 

About "The Qubit Guy's Podcast"

Hosted by The Qubit Guy (Yuval Boger, our Chief Marketing Officer), the podcast hosts thought leaders in quantum computing to discuss business and technical questions that impact the quantum computing ecosystem. Our guests provide interesting insights about quantum computer software and algorithm, quantum computer hardware, key applications for quantum computing, market studies of the quantum industry and more.

If you would like to suggest a guest for the podcast, please contact us.

Start Creating Quantum Software Without Limits

contact us