Transcript
Rene – Hi! Welcome back to season five of QuBites, your bite-sized pieces of quantum computing. My name is Rene from Valorem Reply and today we're going to talk about quantum computing and the market. And I'm honored to have a special expert guest today, Bob Sorensen. Hi Bob and welcome to the show, how are you today?
Bob - Hi Rene! Hey thanks, I appreciate the invitation. I'm really excited about having our chat today. So, thanks for the invitation!
Rene – Yeah, thanks for making it. Can you tell us a little bit about yourself and your background as it relates to quantum computing, physics, math like the whole kind of education, your background a little bit?
Bob – Well, unlike what a lot of people are in the quantum sector I'm not a scientist. I'm an engineer by training and mentality I've been told by many people. So, my area of concentration has always been fast computers. It's like the sports cars of the IT world. So high performance computers, super computers and by natural extension, you know, what quantum computing is capable of doing. Accelerating some very specific parts of key workloads in the in the computing world. So, it was a natural attraction that drew me to quantum computing about five years ago when it really started to gain traction from both the HPC supplier base and the HPC user base. What can quantum do for me from an accelerated computing perspective?
Rene – Right and there's a lot. So let's dive into some of our today's topic and like you mentioned already, you're very much attracted or work in the HPC field since quite a while and your role, you work at Hyperion research as SPP and the Chief Analyst for quantum computing right?
Bob - That's correct, yes.
Rene - Right and so what challenges are actually folks facing in the high performance computing in the high pc sector? What are the challenges that are currently being faced with the classical computing technology or the classical computing architectures we're currently using and how can quantum computing help to overcome these hurdles?
Bob - Well the one thing about the HPC sector is it's never boring. There's always a new challenge and one of the big ones is, you know, for many years users have been able to basically, if they wanted more performance they could sit and wait. Performance in HPCs typically double about once every 14months or so. If you needed a machine that was four times faster, you waited two years and you went out and bought a new one. Well that we're reaching a point where that particular curve of performance improvement is becoming really onerous to continue and I point to an example that some of the fastest high performance computers in the world that are going to be rolling out in the next say six months to 18 months, I'll use the ones that are going in the US, the department of energy government labs, those machines are costing about 600million dollars a-piece now and we're talking about 30 megawatts of power to run them and what that means is these organizations will be spending 30 to 40 million dollars a year just in utility costs to power these systems. So, the projection of where HPC is going, it cannot continue on the current track. We don't see billion-dollar HPC's running hundreds of megawatts of power. So, the world has to change about how designs happen and what we see going forward is the HPC world is going to start to fractionate. It's not going to be one giant machine to rule them all. It's going to be, do I have an AI workload, let me get an AI specific machine. So, I have a modeling and simulation, let me get a computationally intensive HPC. It's smaller, it's less expensive but because it doesn't have to do everything, it'll be more effective for the job it's doing. Which means the sector's moving towards workload specific architectures, which means that this is a perfect time for quantum to come in and be seen as an accelerator for key workloads. So now I have a system and I say, okay, I have a certain workload here. Whether it be machine learning, cybersecurity or optimization and I'm going to build a machine that is specifically suited to those technical requirements, and I'll be able to hang off of that what I would call a quantum accelerator not a quantum computer but an accelerator that for certain workloads can give me performance that is otherwise not possible on a classical counterpart. So really, it's a perfect confluence of interest where the HPC sector in its traditional sense is facing some significant challenges in keeping its performance going. So, it's facing this new reality, quantum is able to step in and say, hey, I can add to the arsenal of computational capability, to your workloads to get you, getting your jobs done faster and better and more effectively.
Rene – Gotcha. And I love what you're saying. Especially that quantum is basically or should be seen as more as an accelerator for specific problems, right? It's also what I keep on reiterating and like in the show but also when talking with different clients but, you know, also folks, that are interested in quantum computing. Some always ask, like oh when will quantum computer replace my computer or my smartphone? And things like that and I always keep on saying, it's like think more about it like a GPU, like a graphics processing unit, which is an accelerator for a specific task. In our case, it started with 3D computer graphics and then of course with these large matrices and linear algebra, they can process in parallel. Of course it anti-ai workloads and what not but, basically the quantum computer is this kind of specific, as the radar, right, for certain problems. And so I would like to ask you also what kind of impact are you already seeing today with applied quantum computing solutions? You mentioned already certain key sector certain key areas in industries where it's most attractive at the moment already and so where in which kind of areas do you see the most impact already today?
Bob -Well first of all, whenever i give a talk at the end, I always put a little silly quote at the end. Just to just kind of leave them thinking and for a long time I used to have one that basically said, let me go check my email on my quantum computer, will say no one ever. So let's put the entire replacement thing to rest for the for the final time. What we're seeing right now is, you know, quantum computing is still in its nascent stage. It's still in a proof of concept. When I get a call from say a venture capital company and they say how can we calculate return on investment or a competitive advantage for some of these systems, to say well, you have to slow down. We haven't reached that particular point yet. So we're seeing a lot of interesting exploration and so when we go out and talk to end users ands ay do you think quantum's going to help here, some of the most emphatic answers we get from places like, first off the quantum computing sector, they're the largest consumer of their own technology. Simply because they're trying to advance it we've seen lots of interest in cyber security, in the financial space, in pharmaceuticals and other biosciences. Certainly academia is very interested in what goes on out there but it's not as if it's a small section of verticals overly enthusiastic and everyone else doesn't care. We see broad interest oil and gas companies, aerospace advanced manufacturing. Everyone is exploring the potential. Now to me, what the most interesting aspect of quantum right now is, because we're in this noisy intermediate scale quantum system. They're not perfect yet. It's going to take some time. So people are looking for applications that are tolerant of noise and the one that is most tolerant right now is the optimization problems. How can I do a job better now than I can on my classical counterpart system because of the noise and such and the limitations of the hardware. You may not get the perfect answer but you may get a better answer and what people forget sometimes is, well what if oh I'm only getting half percent better optimization. They forget scale. If I'm a large transportation sector or I'm figuring out how to monitor all the flights flying in and out of the United States on a 24 hour period, I want to optimize that. If I get a half percent improvement in that process, if that scales across say thousands or hundreds of thousands of truck miles over millions of trucks over years, that half a percent can translate into significant amount of financial return and competitive advantage. So, I say don't look at the actual benefits yet. Look at its ability to scale and so that's where I think the optimization and some of the modeling and simulation activities and one of the things we've seen most recently is the uptake in cybersecurity as a potential great end use case for quantum.
Rene – Yeah, both being a threat but also an opportunity, right, for quantum communication with a quantum secure channel and all of that stuff. So definitely, also like you said definitely interesting with the quantum security sector but also, I love what you said about the optimization part. This is how we approach it with clients. We look into what are the kind of challenges and workloads that would work very well for these kind of optimization challenges and then basically try to formulate it in a different way. For example, into more of a kubo kind of formulation. So that you can put it in certain quantum algorithms. But of course, you need a problem that is suited for it and exactly like you're saying, we'll never send an email on the quantum computer or anything like that right? So, I love that. I got to steal this, Bob.
Bob - One of my favorite examples of an easily accessible concept with what quantum capability is, Airbus ran a competition a few years ago and one of their use cases was, what's the best way to pack the luggage in the hold of an airbus plane that's sitting at the gate because you want to get it unloaded and loaded as fast as possible. So, it can turn around at the gate but computer scientists know that the packing problem, how to take a bunch of different sized objects and optimally pack them in close space is a very complicated problem and virtually intractable when the number of items you're packing gets large. So Airbus said, well, can quantum address this problem? And again, that's the scaling issue. It's okay, we're putting luggage in the hole in a more effective manner but if you do that on every plane and every turn around think of the advantages now where your plane doesn't have to sit on the runway as long you can carry more luggage because it's back more efficiently. It's a scaling issue and it's a classic computer science problem that quantum is well suited towards addressing. So, to me that's such an interesting and accessible example of the potential payoffs here.
Rene - Should I tell you one thing? Actually, the company that won the problem and the problem you just described; it was actually us. I mean, I was not involved in the project but my colleague Giovanni Pilon, he was leading the team that was exactly you know doing that thing you described, with that it's a great optimization the loading problem and yeah amazing. There are so many things involved. I talked in depth with him in QuBites season 2 episode 3, if you want to recap. I hope everyone else does too because like I say it's one of my favorite accessible examples. Let's continue because you have so much knowledge in the quantum computing world and I want to get your, kind of, impression and opinion about the general health of the ecosystem and the future perspective, right? What are you seeing the biggest blockers at the moment for adoption and what is going to happen in the next couple of years that will accelerate growth? We got to look into your crystal ball a little bit.
Bob – Well, the first one from the suppliers the quantum computing supplier side to me is managing expectations. You have to roll out expectations in a way that you don't have the users thinking, well in a year I will be up and running a fully capable quantum system. It's really more about incremental improvement and so the sector needs to, it may not need to develop the perfect solutions tomorrow, but it needs to be able to demonstrate progress. So, you can draw a straight line and say here's where thing interesting things will happen. We've already seen so many companies, we've interviewed over 400 different quantum computing end users or potential end-users, and their aspirations are pretty modest. They want to be able to do some existing workloads at higher performance levels and explore the potential for new workloads. They're not interested in quantum superiority or million qubit processors those metrics are not as interesting to them as can I get my job done better. So, the expectation issue there, I think is key. Manage those expectations otherwise you're going to you're going to run into situations where people start to lose interest. One of the things about, as I said earlier, the HPC sector is never boring and so what happens in the end user community is, they hear about a technology, a year later they want it running in their data center and two years from now they want it to be boring because they're ready to move onto the next advance. So, there's a certain amount of, okay, I’ve heard about quantum let's get going and so we want to make sure that the sector rolls out in a nice predictable manner that everyone can start to understand. And the key here, though in my mind, is providing the platform where the end users get access to the systems because that's where the innovation is going to come from. That's where people are going to say I've got a great idea, let's see if this works. So, get quantum computers of reasonable capability into the hands of all the potential end users out there and just watch how the innovation flourishes, where end use cases and valuable contributions to R&D and innovation and competitive advantage will come from the end users, who will come up with solutions that were never envisioned by the developers of quantum computing. So, to me that's really a big thing. That and make sure the market doesn't ignite in the sense that again too much funding, too much expectations could lead to a certain amount of lack of interest and so you want slow moderate growth as opposed to an explosion of capability where the expectations just cannot be met by reality.
Rene - We want to have a healthy organic growth and is exactly like you're saying, like the whole over hyping of certain technologies is going to hurt in the end. We see the same thing currently happening with one of my other favorite topics is spatial computing, AR, VR and the whole meta wars and web3 conversations. There are really good substantial pieces there that will for surely stick but at the moment there's so much overhype especially coming out of the crypto camp. I have to say where you see so much overhyping with a couple of things, this will hurt us in the end. And that's why we need to be careful with the expectations we're setting and I love what you're saying about quantum computing, it's the same thing right we got to make sure that people realize there's true substantial benefits we can already achieve today for certain specific problems but it's not a generic solution for everything, right? And so, we should be careful. We should look into the real ROI that can be generated right now and also in the in the near term. And then like you said, you know have substantial organic and healthy funding.
Bob - Now your point see you've got a nice brown beard, my beard is gray and then so I can point back to historical events in basically the late 80s there was this huge emphasis on artificial intelligence breakthroughs and oh yeah, there was a lot of funding going into it and a lot of expectations and none of them were delivered with any degree of you know promptness and the sector literally disappeared and people coined the phrase AI winter because there was a huge freeze and it wasn't a winter, it was almost a glacial epic because it took almost 20years until we started to see nvidia GPUs and some interesting algorithms and deep learning that revitalized the sector which by the way was a completely different flavour of AI than what we were looking at in the late 80s and 90s or early 90s. And so, there is this, we don't want to say it out loud but the idea of a quantum winter is something that people kind of rumble about after you've given your presentation. You've met some friends in the bar and now you're talking over drinks and you're saying, well what do you think and so everyone is very cognizant of the fact that if the growth isn't predictable, if it isn't demonstrated, the spectre of an AI like winter phenomena is always in the back of everyone's mind. So, we do want to avoid that spectre as well.
Rene - And it's good that folks learn from the past right and we try to avoid it and the experts are talking about it and are aware of these challenges but also see the opportunities ahead. Well Bob, we could talk for many more hours. It has been a fantastic conversation but unfortunately, we're already at the end of our short show. We should probably get you back for another episode and continue the conversation. But anyhow, thank you so much for joining us today and sharing your insights. That was very much appreciated.
Bob - Thank you, it was great meeting you and I really enjoyed the conversation and as I said earlier you have made a major contribution to the sector by making this kind of technology accessible it's not all magic. You just have to spend a little extra time and I think you've done a great job, you know furthering that concept, so thank you for inviting me!
Rene - Thank you so much for the kind words as well and well thanks everyone for joining us today for yet another episodes of QuBites, your bite-sized pieces of quantum computing. Watch our blog, follow our social media channels to hear all about the next episodes of season five but also check our website if you want to recap some of the previous episodes. You can find them all there. Until then take care and see you soon, bye!
