Latest / Technology Now / Sixty years of innovation: how would we use a quantum computer? | Masoud Mohseni
Transcript
- MASOUD MOHSENI there has been one thing constant that at any point if you ask a quantum competing researcher, when we will have, quantum computer, that will tell you it's 10 years away. You would get the same answer back in early 2000, and you get almost the same answer today, but there is a difference. So, at that time, 10 years was, the shortest time, it was the lower bound. but when people talk about 10 years today, that's more of a upper bound, we don't expect to wait more than 10 years. SAM JARRELL So the answer is the same, but the context around its meaning has changed? MICHAEL BIRD Yeah it’s an interesting one isn’t it – and when you hear that clip from my interview with Dr Masoud Mohseni later, it’ll probably make a little bit more sense in context. SAM JARRELL But we are returning to the nice, simple, easy to understand topic of quantum computing today? MICHAEL BIRD We absolutely are. I’m Michael Bird. SAM JARRELL I'm Sam Jarrell MICHAEL BIRD And welcome to Technology Now from HPE. MICHAEL BIRD All right, well welcome to the first of five themed episodes we are making in collaboration with HPE Labs to celebrate 60 years of innovation. SAM JARRELL That's right. We'll be releasing these episodes throughout the year, so make sure to keep an ear out for them. We are going to be exploring some of the most interesting and influential research that has happened and is currently happening. At HPE labs and how this could impact all of us: businesses, organizations, and everyday people. MICHAEL BIRD Yep. And for this our inaugural episode, we are going to be diving at the deep end with, as you probably realized, quantum computing. Now, a few months ago we did a sort of quantum computing 1-0-1, which we'll link to in the description. So today we are looking not at how the computers physically work, but more importantly, how they can be used in practice SAM JARRELL Because these things aren't cheap, right? I can't imagine there's much point in building a quantum computer if you can't actually link it up to your regular infrastructure and use it MICHAEL BIRD Yes, Sam, very much so. And figuring out how to effectively integrate quantum computing into high performance computing or AI is something that is being studied extensively at HPE labs. So to find out more about how a quantum computer could be integrated into our modern day infrastructure, as well as the newly formed Quantum Scaling Alliance, I spoke to Dr Masoud Mohseni. Masoud is a Senior Distinguished Technologist, and Director of HPE Quantum at HPE Labs and before we got into the weeds of discussing quantum computers themselves, I wanted to know a little bit more about Masoud and what attracted him to studying quantum computing in the first place. MASOUD MOHSENI I got into it more than 20 years ago, early 2000. I was a PhD student and, I was driven by this kind of question of what would be the ultimate limits of computations and, learning based on the laws of physics. So I just want to push the limits of computation. To the extreme limits that can be done under the norm laws of physics. I was fascinated by, studying complex systems in both classical and quantum regime. given the, everything that was going on at the time, quantum computing was just beginning, I thought that this is a very interesting topics to see if we can use fundamental laws of physics at the quantum scale to accelerate computation and also improve our modelling. MICHAEL BIRD So you've been doing this for about 20 years. Um, how long has, how long have HPE labs been working on? quantum computing, quantum computers, MASOUD MOHSENI I came to HP three years ago and, after I joined HP that was a transition in the HP, studied quantum competing before I arrived. but, I basically, changed, the direction of research and we went for kind of a full stack, quantum HPC integration as, Main, mission statement for, quantum team. after I joined hb, we went to kind of a. designing a full, blue blueprint for how to build quantum supercomputers. MICHAEL BIRD so you've been doing research on quantum computers for the last 20 years, as you said. how has the technology moved in those 20 years? MASOUD MOHSENI Yeah, that's a very good question actually. there has been, many changes, over the past 20 years. but there has been one thing constant that at any point if you ask a quantum competing researcher, when we will have, quantum computer, that will tell you it's 10 years away. You would get the same answer back in early 2000, and you get almost the same answer today, but there is a difference. So, at that time, 10 years was the, the shortest time, like that, that it was the lower bound. but when people talk about 10 years today, that's more of a upper bound, we don't expect to wait more than 10 years. It has, it will happen less than that. So this is a big shift in the expectation but having put this aside, the research was at, back in the day, more theoretical in the algorithm and error correction But then there was a transition, around 10 years, 12 years ago that this becomes a major industry kind of research. MICHAEL BIRD And so, why are we developing quantum computers? I mean, we've got perfectly good, classical computers at the moment. Like what will a quantum computer do that classical computers at the moment can't do. MASOUD MOHSENI obviously we had Moore's Law and We have reached certain capabilities in ai, But I should tell you that certain problems are much bigger and harder to compute. So you can think about GPUs accelerating linear algebra calculations, uh, but they do it let's say by a factor of, you know, a hundred. But there are matrices and vectors that are exponentially big. So quantum system live in, a hilbert space. This is a mathematical space that represent our knowledge about quantum system that is exponentially big. Let's say if you add any single qubit. You have to double up the amount of classical compute. And just after around 50 qubits, as you approach 60, you are gonna run out of memory of compute time, And it's impossible to scale it up. So you can think about quantum processing units or QPUs as hardware accelerators for these exponentially large matrices and vectors, MICHAEL BIRD And, are there any problems that a quantum computer will be uniquely good at? what are the sort of problems and tasks that a quantum computer would, would absolutely blow a classical computer out the water with? MASOUD MOHSENI It's, uh, very naive to think that you are gonna use, Quantum computers for every workload that you have today. generally speaking, they're gonna be very powerful machine for things that are inherently your processing quantum data or your modeling quantum data. you might say, what is quantum data? It's, our knowledge about molecular system. when you want to simulate quantum chemistry calculations, you want to study new materials, new state of matter, new drugs You are dealing with quantum data and those are gonna sit beyond an exponential wall for classical computers as we know it. MICHAEL BIRD what sort of industries will, be the first to, acquire and use quantum computing? MASOUD MOHSENI Yeah. So for the application Would be the industries that would have to come up with new materials, pharmaceutical companies, automaker, for new batteries, new solar cells, when you wanna study, nuclear fusion, So if you model them or try to understand them purely classical, you would have very poor model or calculations, would blow up exponentially. MICHAEL BIRD and you mentioned error correction. Is this where physical qubits and logical qubits come in? MASOUD MOHSENI Yes. So people usually discuss, physical qubits because they're larger. but usually there's, it depends on the code you use. There's a 10x overhead, going from physical to logical qubits, logical qubits. Why you need them? Because you need to introduce redundancy such that the noise can be detected and corrected, So basically you encode one bit, let's say, into, three bits, like a repetition code. And this is. Known in classical, competing as well, but for quantum it's more complicated because you have to, correct for, not only this kind of classical noise, but there are quantum noise because you're kind of messing up the quantum coherence or phase of Quantum computer, and that's harder to detect, MICHAEL BIRD , so a logical qubit is made of many physical qubits, and those physical qubits are working on the same problem, and I'm obviously oversimplifying it here, but, between those, say 10 physical qubits, you basically take like a, an average or a vote as to which answer is correct amongst them. Have I got that correct? MASOUD MOHSENI Yeah. The entire problem with quantum computers is that they're very prone to noise than classical computers. It just because we encode the information in quantum phases that very susceptible, they interact with environment strongly and that -has been the biggest challenge to be quantum computers. MICHAEL BIRD So, we've talked about some of the applications for quantum computers and, they all sound like the sort of problems that, People have said, AI is gonna be fantastic at these sorts of things. So would we run AI on a quantum computer? MASOUD MOHSENI So I think what we are gonna face in future is that, quantum and AI are gonna be integrated, and being, basically, complimentary approaches to simulate and model, They're not gonna be competing, Basically a three-way integration is happening is, there is an integration of quantum with HPC But then AI has to be integrated with, HPC, systems there are certain, kind of contribution from AI at the lower layer of, uh, quantum supercomputer, You need AI for, Lower layer of a stack to help with the real time, calibration, real time control, and real time error correction. You also need it for workload management at the higher layer of a stack, though. for example, for lead, drug, candidates where, you have certain, chemical compounds that you don't know if they're good or bad, There are ways to either simulate this, or do some experiments in the lab. Experiment is very hard. could take years. It's very expensive. So you, ideally you wanna actually simulate them and exclude some of the bad ones But the simulations are exponentially hard, so you can run those simulation. On quantum computers. and generate high quality synthetic data, and that could be fed into a classical ML pipeline. so at the higher layer of the stack, actually quantum competitors can enable even classical ml. MICHAEL BIRD this concept of, a quantum computer being an accelerator for a high performance computer, is that what we're talking here? MASOUD MOHSENI Yes, exactly. if you look at what would be the first application of quantum computers, this kinda workload are being currently run at some of the DOE national labs and various supercomputers that most of them actually, are. Produced by HPE. But some of the workload, you say this sub problem is just too quantum. so I wanna offload it to a quantum computer to have like a interplay. So it's in the current pipeline. It's much easier to start, bringing quantum computers to prime time by deploying them on the current HPC infrastructure. And that's the reason why HPE Labs is investing in this. we believe that we can do this in much lower cost in, r and d and lower cost of manufacturing, but just upgrading our existing supercomputer to become quantum enables by adding quantum ranks. MICHAEL BIRD All right. So, just to change tact slightly. you lead the Quantum Scaling Alliance. Can you just talk me through what the Quantum Scale Scaling Alliance is? MASOUD MOHSENI So Quantum Scaling Alliance, Is founded by, HPE and basically I co-founded this with John Martinez, who as a novel laureate, was actually, awarded Nobel Prize for his, pioneering work on quantum superconducting processors, in the study microscopic quantum states. so we had this vision of, building a. Consortium to build quantum computers by a horizontal integration. So usually in various different comp companies that pushing to build a quantum computers, they have this vertical integration that they're building from hardware and, different layers of software and even application at the top to just like, offer quantum solutions. We believe that, horizontal integration is better because once you are in a single company, if you make mistakes and you choose a particular QP modality for processing, usually it's very hard to, pivot and change, the course. If there are like, evidence that. Particular modality might not be the best. This horizontal integration has long history in, conventional semiconductor technology. This has been, the normal form of operation for decades. So it's just, there is not a single company offering the full stack. you are basically agnostic to underlying QPCs, And you can, pivot and choose a different solution or breakthrough when they become available. this reduces the cost, minimizes the risk, and Most of this would be, open source and open architecture. MICHAEL BIRD and have you discovered anything in the months, since the Quantum Scale Alliance was officially formed, that you're allowed to tell us about? MASOUD MOHSENI we had the, extraordinary demand and like various companies want to join. we are setting up some, governing, and regulations for, membership and how to operate. So one of the things that led to this consortium was we wanted to write a position paper we call it. But it ended up as we did a lot of research to kind of, outline a roadmap, for how to build, quantum supercomputers. for that kind of work, we started by listing all the current challenges to scale quantum computers from hundreds of KBIs to millions. And, we Try to come, clean and mention all the challenges that are, underappreciated or were untold. So, to be able to work as a consortium and, try to provide solutions. And it seems that, since we have done that, there is a very high demand for doing the same, for even different modalities like. we provided the blueprints for, quantum computers based on superconducting cubits. But you can do the same for, let's say, neutral atoms or trapped ions, or a spin based cubits. these are, top tier candidates And so there is a kind of a good traction across the industry for, Identifying the challenges and, mentioning them publicly rather than staying close within a particular, company or, startup. we cannot solve problems that we have not, acknowledge their existence. MICHAEL BIRD How long do you think it will be until we have a functioning useful quantum computer? MASOUD MOHSENI it depends, what do we mean by, functional or practical computers? my definition would be, quantum computers relevant for. practical application when the news of solving something on quantum computer, that the solution that it provides is more important that it was using quantum computer to arrive at the solution. So I think that kind of applications, you could say. Could be as for us, like five years to 10 years. It just based on the progress. today, if we get to tens of thousands of physical qubits, which corresponds to a few hundred logical qubits, I can see practical application. that in, in some form of, quantum chemistry calculation on RSA factory that could become possible. MICHAEL BIRD I could keep asking you questions about quantum computers. maybe we'll follow up with another episode, but thank you so much for your time. It's been an absolute pleasure speaking to you on, uh, on technology now. MASOUD MOHSENI Yeah. it's been a pleasure, having this chat with you and, thanks for invitation. SAM JARRELL I think it's fascinating, that he already brought up the, the time horizon of like the 10 years. but then even like the useful applications are only about. Five-ish years out as well. so to me, this is like very much an inevitability. but I still think that people themselves are gonna be caught up in exactly what he sort of flagged. did a quantum computer do it or did it make sense to actually use a quantum computer? This is the same issue we have with AI right now, right? Where it's like AI for the sake of AI. MICHAEL BIRD Yeah. And it feels like there's so much talk about quantum computing in the last. Six months or so. I mean, certainly we've, we've talked about it quite a lot on the show. It feels like momentum is building, quantum computing doesn't just feel like this thing that's, it'll be nice and the researchers will use it. It's going to impact us in our daily lives. we've talked about post quantum cryptography and here we're talking about actually, drug discovery and things like that. So. It feels like momentum is building. SAM JARRELL I think, there's probably like cynicism. Across like industries but I think people probably underestimate how hard it is when they hear like the qubit counts. Like the fact that you might need 10 physical qubits for just one logical qubit purely to deal with noise shows. Why, like the scaling, it's a massive challenge, right? That's why there's things like the Quantum Scaling Alliance, right? I understand too, the frustration of hearing it's 10 years away. It's 10 years away. but that 10 years feels, very inevitable Now. MICHAEL BIRD Yeah. Yeah. I'm glad you said inevitable because I, I liked the, quantum AI and high performance computer in integrated. It sort of feels like it's a sort of a superhero, team up, doesn't it? So, um, like, it sort of feels like that we're, we're gonna have quantum AI and high performance computer and they're, we're gonna work together and we're gonna have this, we're gonna be solving some awesome problems with it. I mean, do you think we will have like our AI moment when AI hit the mainstream and then it just accelerated from there. Do you think we're gonna have that with quantum computing SAM JARRELL I think it depends. 'cause this isn't like, a new laptop, right? So there's not really like everyday consumer grade quantum computers I think that are going to be available for people the way that LLMs are. But I think if like. There are large commonly known organizations that start talking about they're using it and then what are the outcomes that they have. MICHAEL BIRD I feel like I'm slowly wrapping my head around a quantum computer and why we need physical qubits and logical qubits. I mean, understanding it from like an error correction perspective is starting to make sense to me. and now I sort of understand why this is becoming such a complicated problem. We may already have qubits, but actually you'll need thousands, tens of thousands, millions of qubits to actually do something useful with it. SAM JARRELL I honestly, I appreciated learning how open. The work on this is because I didn't realize that it was sort of like open source to some degree, to solve some of these hard problems. To me, when I think of like quantum computing, I think of something like, I don't know, very secretive to some degree. MICHAEL BIRD I agree with you and I really actually enjoyed the conversation, around. how quantum and AI potentially will work together. because I really struggled to wrap my head around like, well the things that we say that Quantum's good at is also things that we sort of said that AI's good at. So play to the quantum strengths, like the sort of simulations, the chemistry bits of it. and then play to AI strengths. So, we talked about, using it to synthesize data, and so actually they can work in partnership to do things a bit faster. SAM JARRELL they're complimentary, not competitive. And so quantum helps basically generate high quality data for those problems. And then AI sort of helps with the control and calibration and maybe even deciding which problems actually warrant. Using a quantum level system, versus just your traditional setup, which you know, is still very powerful. MICHAEL BIRD I think you're right. Now, Sam, while Masoud mentioned the length of time it could take until we have a useful quantum computer, which could be used in industry. One thing we didn't cover in much detail was the negative consequences to our security that could also be realized. So I wanted to finish off by asking Masoud a question we ask of all of our quantum guests, how long will it be until we have a. Cryptographically relevant quantum computer and his answer will bring us back in a nice closed loop. If you can still remember the start of the episode. MASOUD MOHSENI People usually assume that that's gonna be needing tens of millions of qubits for RSA integer factoring, let's say 2048 bits integer. But there are certain protocols that intes that you need to factor is a smaller. And so that number, reduced, in a series of publication over a past year to under a million, and then, some claims of around a 100k physical qubits. I should mention that there are some assumptions in those studies. Where, the noise models are not realistic, so it might be a little bit too optimistic, but if we have, with hundred thousands of physical qubits, it's not impossible to think about crypto graphic applications and that could happen within next five to 10 years in the optimistic scenario. SAM JARRELL Okay that brings us to the end of Technology Now for this week. Thank you to our guest, Dr Masoud Mohseni And of course, to our listeners. Thank you so much for joining us. MICHAEL BIRD If you’ve enjoyed this episode, please do let us know – rate and review us wherever you listen to episodes and if you want to get in contact with us, send us an email to technology now AT hpe.com and don’t forget to subscribe so you can listen first every week. Technology Now is hosted by Sam Jarrell and myself, Michael Bird This episode was produced by Harry Lampert and Izzie Clarke with production support from Alysha Kempson-Taylor, Beckie Bird, Nicola McCombie, Alissa Mitry, and Je nessa Ayache. Our theme music was composed by Greg Hooper. SAM JARRELL Our social editorial team is Rebecca Wissinger, Judy-Anne Goldman and Jacqueline Green and our social media designers are Alejandra Garcia, and Ambar Maldonado. MICHAEL BIRD Technology Now is a Fresh Air Production for Hewlett Packard Enterprise. (and) we’ll see you next week. Cheers! SAM JARRELL Bye y’all