Martin Myers: In a post quantum world, the cryptography that we’re using today is virtually useless. Everything that you are communicating via your browser today, will be readable in the future and anything you've done today that's recorded will be readable in the future as well.
Michael Bird: Hello and welcome to Explain IT, brought to you by Softcat; the show for IT professionals, by IT professionals that aims to simplify the complex and often overcomplicated bits of Enterprise IT without compromising on detail. I'm host Michael Bird and over the next 30 or so minutes I'll be challenging our panel of experts to take a different area of the IT ecosystem and of course, explain it. In this episode we’re going to be talking about quantum computing; what on earth it is and how organisations might be able to take advantage of it in the future and what the future might look like for quantum computing. With me to help is Craig Lodzinski who is Softcat’s Chief Technologist for data and emerging technology. You'll be very aware now that we ask all of our guests to bring along an interesting fact. So your fifth interesting fact please Craig?
Craig Lodzinski: So I think, we were having a bit of a motorsport conversation before we started and that spurred my memory to remember that I once had seven times Le Mans champion Tom Kristensen make me a cup of coffee.
Michael Bird: Ooh lovely! Why did he make you a cup of coffee?
Craig Lodzinski: Before I got into IT I used to work in events, and we were doing a motorsport event and it was absolutely hammering down with rain outside and I was doing a bit of marshalling and Tom was kind enough to bring some coffees out of his hospitality unit for us.
Michael Bird: And we’ve also got Martin Myers who is Fujitsu’s CTO for the UK product business. Martin, as you heard we ask all of our guests to bring an interesting fact. What is your interesting fact?
Martin Myers: When my daughter completed her A Levels, she’d been studying geology, and I thought to be a great idea to take her to Stromboli, which is a volcano within the Mediterranean, and we arrived there and we climbed to the top of the volcano and we looked down into the burning cauldron of lava spitting out and being quite obnoxious. As a result of which, she is now studying biology at university. Not geology.
Michael Bird: So let's jump into the show. Craig, Martin, what is quantum computing?
Martin Myers: Quantum computing today is an emerging field. We're not yet fully developed in this area. It's really exciting, we’ll be able to really do some cool stuff with it, resolve some really complex problems. It's going to be fantastically expensive to get to the point of commercial utilisation. But the most important thing is it's not an evolution of the current IT environment that we know today, it uses some quite complex quantum-mechanical phenomena.
Michael Bird: So can you just quickly explain then, some of that quantum mechanical phenomenon?
Martin Myers: So there's three main phenomena that quantum computing is going to utilise in one form or another and there’s various different types of strategies being adopted. The first of those is a principal called quantum entanglement and this is the ability to link the properties of two particles and then separate them to some great distance while maintaining the link properties. Affecting the property of one then affects the property of the other, even though you may have separated them by thousands of miles. You might recall that Einstein referred to this as spooky action at distance cos he was quite dismissive of this principle. But it has been shown mathematically to be correct now and experimentally as well.
Michael Bird: Where have they experimented with it?
Martin Myers: It’s been done in various research laboratories and in a few commercial organisations as well, to the extent also that the Chinese have recently launched a satellite and demonstrated that you can entangle particles on earth with one of them out into space and maintained a coherency between the two.
Michael Bird: Is that instant?
Martin Myers: It is instant.
Michael Bird: Or is it speed of light?
Martin Myers: It is instantaneous.
Michael Bird: So what’s the second phenomena then?
Martin Myers: So the second one is equally bizarre - it's the principle of quantum superposition. So everybody will be familiar in today’s IT environment that we use bits which represent a zero or a one, and that's the maths method that's used in all computing today. With quantum superposition, the qubit, essentially the equivalent to the bit in the quantum computer, that would be representing a zero, a one, or any other value in between as well. So you don't directly know what the value of that is until you actually evaluate it at the end of the calculation. In the meantime it can represent a zero, a one, or any value in between. Michael Bird: Wonderfully confusing. And so what's the third point?
Martin Myers: So the third principle is the principle of quantum tunnelling, and this is another quite strange phenomena. This is the property that a subatomic particle may transition from one energy state to a lower energy state even though it needs to pass through a higher energy state to get there. In a real-world analogy that will be like an orange sitting in a vase, passing through the side of the vase onto the table to get to a lower energy position. Now theoretically that's possible but the probability is extremely low.
Michael Bird: Ok so these three principles don't sound particularly conducive to good computing, especially the quantum superposition about the binary state that can be in a zero or one, can you just explain a bit about quantum computing, how that works and how that uses all of these quantum phenomena.
Martin Myers: So these are ideas that were first developed in the 1980s and there were several different researchers and different research teams involved in this across the world, but one chap in particular, a chap called Feynman, he did a lot of the original work in the space and it was recognised to the extent that he got a Nobel physics prize for that. Using those quantum phenomena is extremely difficult to do and as I alluded to earlier we use qubits rather than bits and we represent numbers in a way that can be nought, one, or any number in between in parallel across the entire workspace at the same time. And then what happens at the end of the process, which happens very quickly, is we get a probabilistic answer, so unlike a classical computer which tells you the answer is 3.1415 for pi, the quantum computer will tell you that it's probably 3.1415 with all the other digits filled in, with a certain level of certainty and then it's up to the scientist then to evaluate if that gives it that level of certainty that is required for the particular task in hand.
Craig Lodzinski: So as Martin described, one of the key factors in using particularly quantum superposition for computational purposes is that instead of a bit being binary, being zero or one, a qubit can be a quantum superposition of that zero or one, so that gives you a continuous string of variables between those integers. What that means is that, so in a traditional system if you take a classical CPU with a million transistors, you add a single transistor that's one part per million improvement. If you take a five qubit system and add another qubit, you’re doubling the computational power of that because you have an exponential function in how you’re able to take that continuous string of variables and perform computation on it. So that makes the potential of quantum computers very powerful.
Michael Bird: So what would you be able to do with a quantum computer that you can't do with my laptop that’s sat in front of me then?
Martin Myers: So that's a really good question because I think a lot of the activity that's going to be in the space of quantum computing isn't going to be directly related to the traditional IT and ERP systems that we’re used to dealing with today on a day-to-day basis. Whereas IT and office automation using in the digital world is quite well developed, quantum computing is moving into a space where you're looking at problems rather than automation and one of the aspects of that is that in IT today, we can process an awful lot of information. There’s certain types of problems that we can't address, those are difficult problems, they’re combinatorial problems and those fall into many different fields. Typically today where those problems occur, we drop them out into a human resource and the human evaluates the task and then we go back into a compute world. So I think the biggest effective quantum computing is going to be in those spaces where today we’re using an expensive organic resource, a human, to undertake a task and then transform that into a quantum environment, it will not speed up the typical office automation tasks that we are doing today in a routine basis.
Craig Lodzinski: It's very important to stress that quantum computers are very very different to classical computers and it's not going to be a case of we just switch over. There’s the idea of quantum supremacy, which is when quantum computers become viable, but that's really the base level in terms of it’s where we can't simulate quantum in terms of a classical computer. So if you take, for example, very deep chemistry. To take a 50 to 60 atom molecule, simulate all the electrical functions within that, how that behaves, is very difficult with a classical computer because you are applying effectively Newtonian physics in the way we perceive the world and a binary system in a classical computer to something that obeys the laws of quantum mechanics. So to calculate the Schrodinger equation, which is how you figure out the quantum mechanical effect on the set system, is very difficult in classical computing as you scale that out because there's a problem, and we talk about Moore's Law being the reason that computing in a classical sense is slowing down, there’s also a concept called Amdahl’s Law, which effectively states that you have diminishing marginal returns as you parallelise the system. There’s certain things that have to be done in serial, and quantum computing is going to start to address those types of problems and certainly when you think calculating the Schrodinger equation for very large molecules that’s going to really change how we deal with chemistry and it may in fact be that the first quantum computers are used to perform material science to figure out how to make a better quantum computer. Talk about qubits, in classical computing we have the bits which is logical and we have transistors which are the physical game. In quantum computing qubit is the physical and the logical element, because we haven't yet had that watershed moment - in the 1950s we figured if I take a load of transistors, put them on a piece of silicon and that became a chip -we don't have a defined standard for a quantum integrated circuit yet. We're starting to get there, but because of the nature of the quantum computer it may well be that the first quantum computers are used to perform quantum chemistry that help us build better quantum computers that we then start to accelerate on.
Martin Myers: So two important points that Craig mentioned there. The first is that actually we can't do this today, it's extremely difficult. We've got some small scale demonstrations in lab environments of quantum computing, but there's no commercially available systems.
Michael Bird: And why is that?
Martin Myers: Because of the complexity and the technical challenges in establishing a sufficiently stable environment and within which to create the quantum environment. It's a physics challenge which hasn't yet been overcome.
Michael Bird: And what is that challenge?
Martin Myers: It's to do with establishing a thermal stability and we're talking in the order of millikelvins that you have to take the temperature down to, with zero vibration, zero noise, shielded from a magnetic environment as well, it’s quite a complex task to set up because a lot of the machinery you need to create a millikelvin environment creates vibration; compressors, pumps etc, and electrical noise, so there’s quite a complex task that needs to be overcome. And the second point there was that Craig was talking about some of the aspects where quantum computing will likely be addressed in future and those aren't the typical calculations that we do with computers today. You would address them with a different way because it's a different type of compute device. And very much of what I'm trying to get across here is that if you've got a task that you want to speed up on today's current computer, don't wait for a quantum computer to come and do that for you because there's probably a better way and that's probably as simple as just doing an upgrade; faster processors or a newer server, parallelisation or application optimisation, those would be techniques you could use to improve performance on today’s compute, whereas quantum compute will address a different workspace in the future. So in terms of parallelisation and application optimisation, those are all things that can be done today. Quantum computing, as Craig said, it's several years away in terms of a commercially exploitable platform. But what it's going to do is quite a different task, as Craig said, and here’s a classic computer problem which illustrates the kind of benefit that quantum compute is going to give, and it's called the Travelling Salesperson Problem, and a lot of people will have heard of it, and the problem is around how do you route a salesperson around his customers in the most efficient or shortest fashion? And it's quite a simple task to solve if you've got five customers there's only 60 possible combinations of going from A to B or A to C or A to D and then to B, A and back again. However the problem gets more complex very rapidly. If you had 10 customers, that increases to 1.8 million possible combinations of routes, and if you had 30 customers, then it's a very large number it's 1.3 with 32 zeros after it which is, I’m not even going to try to pronounce that right here. Now that kind of task becomes very quickly virtually impossible on a current computer but it would be a trivial task on a quantum computer and could be resolved in a small fraction of a second, a few millionths or a billionth of a second. It doesn't solve the problem though, that most salesmen are going to want to route their visits around the brasserie at lunchtime, the golf course in the evening; those tasks will have to be worked into the quantum compute environment as well, there needs to be a bit of machine learning perhaps. And there's plenty of other examples across industry where we’re doing a similar kind of thing, for example the iso containers that you see on the back of lorries or on ships, there’s 20 million of those in use at any one time, probably 200 million different shipping options, per container, so the combinatorial task of working out what's the optimum route of shipping those around the world is beyond any form of classic computer today. Production line scheduling is becoming increasingly important with just-in-time delivery optimisations, so how do you ensure that you've got the right products in time to deliver just in time on to the production line to guarantee your deliveries becomes increasingly important, that is a combinatorial problem. Establishing the optimum routing for efficiency of robots in a production line is also quite a complex task and especially when the robots interact with each other and working in the same model space.
Craig Lodzinski: That broad category of optimisation problems is something where quantum computing sits naturally. You look at the type of mathematical problems that Martin suggests there, like the Travelling Salesman Problem, as I said previously, the way that quantum computers are structured and how you get that continuous variable and that exponential rise leads itself to these exponential type of functions, so certainly there will be a lot of use cases where you can take optimisation issues because of the difference in quantum and classical computers you may have these two working in parallel. They’re very strong parallel system for neuromorphic computing, FPGA computing, GPGPU computing for things AI and machine learning. Figuring out what algorithms, what training sets to use on that, actually taking probabilistic computing and quantum computing, trialling different algorithms, figuring out how they work mathematically and then recommending that back to an AIML cluster can help to iterate that a lot more rapidly.
Michael Bird: Ok so for organisations, can they go out and buy a quantum computer today?
Craig Lodzinski: No.
Michael Bird: And why not?
Craig Lodzinski: It's very very complicated stuff and there's a number of steps we’re going to take along the way to get to quantum computing, it's a completely different concept. At the moment there's still plenty of optimisation to be done in classical computing. Typically we’re at 10 nanometre process node for current CPUs. There is movement to get that down to seven and five on the road map. Below five we're going to struggle and this is where Moore's Law starts to tail off and we’ve kept that going with parallelisation and multicore units, but increasingly we’re going to see more specialised systems, so the emergence of FPGAs and specific ASICs and directly programming chips in order to suit certain applications and we're going to get certain improvements in CPU technology, but it’s starting to slow down, not only because of the physical limitations, so actually as we start to shrink further, we start to come across that quantum tunnelling effect because we’re building transistors already that are a thousand times smaller than the width of the human hair, making them smaller again starts to come into all sorts of crazy quantum effects because you get so small, and you often have a lot of heating and cooling problems, material science issues as well. So we have that bridge in widening out away from the traditional X86 orthodoxy, but we’re also bringing in specific types of chips so neuromorphic computing, FPGA specific chips and also Fujitsu have got a product called digital annealer, Martin.
Martin Myers: So digital annealer, it’s really our bridge towards quantum computing. So this is a service which is offered out of Japan and it's going to be a physical offering in the UK as well. And what we've done, the scientists in Japan have looked at what a quantum computer would do with quantum tunnelling in the future, when it would actually be available and they’ve analysed the maths of what that quantum computer would do, and then implemented that maths into silicon, in today's technology. So we have now got a PCIE card with a digital annealer chip on it which we can start to simulate what a quantum computer would do in 10 years time. Now that ASIC that is running the digital annealer is nothing like as quick as a quantum computer, but we’ve wrapped the same software environment around it that's being developed by the likes of D-Wave or 1Qbit, so we can start to simulate what's the look and feel of a quantum computer would be and start to perform some of the calculations as well.
Michael Bird: So for an organisation, what would they use a quantum computer for? Presumably they're not going to be running CRM systems from it. What are the main advantages?
Martin Myers: You are absolutely right, the CRM systems, ERP systems, any kind of office automation that we’re doing in today's classical compute, I think will probably remain closely in those domains as well. This is very much, quantum computing is very much looking at space where the calculation that's required and it's looking at a problem, something that you can calculate, rather than something that you can automate - I think that's possibly the distinction between the two, but the calculations that are performed today by organic resources, usually expensive organic resources, so either scientists or portfolio managers or some kind of manager of an organisation who’s trying to evaluate how they can best optimise the routing of robots or production lines or logistics etc, the digital annealer and the quantum computer will directly address those kinds of spaces rather than the office automation spaces that we’re familiar with today.
Craig Lodzinski: Whilst we’re trying not to make hard and fast predictions on quantum computing in this episode because they're bound to be wrong, I can categorically say you won't be able to play Crisis on it, it’s not going to be used for general purpose computing applications because of the nature of what quantum computers are useful for. Optimisation problems are going to be the big key in terms of a broad field of study. Any sort of optimisation problem for greater efficiency for optimising a really broad function is ideally suited to the use of quantum computers. We’ll see quantum chemistry and also we’re going to have a very interesting field which is something that's really spurred on a lot of the investment, particularly from governments, which is called quantum cryptography.
Michael Bird: That is probably the coolest name for something I’ve heard.
Craig Lodzinski: I'm assuming in 10 to 15 years’ time all the people who are leading edge blockchain philosophers and first tech startup VC guys in blockchain, they will all become quantum cryptography specialists.
Martin Myers: That introduces an interesting point there, because in the post quantum world, the cryptography that we are using today is virtually useless. It’s a trivial task for a quantum computer using shuttles algorithm, and the algorithm’s already been written, to do the prime factorisation challenge which is a core to the asymmetric key infrastructure that we use today.
Michael Bird: So what does that mean in layman's terms?
Martin Myers: Anything that you are communicating via your browser today will be readable in the future and anything that you've done today that's recorded will be readable in the future as well.
Michael Bird: So files that are encrypted today, if somebody manages to, in secret, develop a quantum computer and it's really powerful, or it does everything it supposed to, could that person then take some encrypted files and just decrypt them without anyone realising?
Martin Myers: Yes, if they're encrypted via asymmetric-key then yes.
Craig Lodzinski: There's a really big reason why there's a lot of military and government funding going into this and this is the reason.
Martin Myers: So in the US they are evaluating a number of different algorithms and a number of different techniques that they want to put forward and test for a post-quantum world. And in a year or two's time I think we might have a slightly different cryptographic environment.
Michael Bird: So what you're saying there is that they are already developing the next levels of cryptography, but using quantum computing?
Martin Myers: Well they’re developing the next generation of cryptography to accommodate the advances in quantum computing to render those next steps in cryptography to be proof against quantum computing. Everything we’ve done in IT, in all of our industries has been evolution of what came earlier. This is completely different, it's a revolution, it's not an evolution of anything.
Michael Bird: I guess what you're saying is, whatever we have learnt, whatever we've developed and built in the last 10, 30, 40 years is basically…
Martin Myers: 100
Michael Bird: 100 years, is basically useless because it doesn't work in the world of quantum computing.
Craig Lodzinski: Everything we’ve built so far is based on the Newtonian model of physics. Everything is underlined by that because it's based on what we perceive we see, because it's been developed before we've really understood the quantum-mechanical side of things. Now we understand quantum mechanics, we’ve now got to rebuild the fundamental model of how we deal with everything in order to take advantage of it.
Michael Bird: So we talked about, I guess, some of the stuff that is being looked at the moment, algorithms that are being developed, the digital annealer we talked about, but when can I go online and buy a truly quantum computer?
Craig Lodzinski: The short answer is we don't know. Longer answer is there's a lot of steps to be taken to make that a reality. It's not going to be all that soon. If you are Softcat and Fujitsu customer and you're looking at saying maybe we hold off on our hardware refresh for a year or two because we're not going to be using conventional Xeon powered servers in two years, we’ll be using quantum computers, and this is not a sales pitch, but please come and buy some more servers from us because we're not going to be able to sell you a quantum computer anytime soon, as much as we would love to.
Martin Myers: And it won't address the workload you’re doing on a classic computer anyway.
Craig Lodzinski: Absolutely yeah. I think the field’s going to be very similar to what we saw in classical computing. The ‘canary the coal mine’ is going to be the government and military application of these computers. Once we start to see universities, governments, research institutions all coming together and we're starting to see some really interesting papers come out, but once we get these things breaking ground and papers coming out saying we've done it, you’re then going to have the same type of lag that we saw, perhaps not to the same extent as we saw with classical computing because of the limitations of when we discovered classical computing in a real commercialised aspect from the Second World War the Cold War, but there's definitely going to be that lag between the research and government applications and cracking that and then the shift wherein IBM, Fujitsu, Google, Microsoft, Amazon will be able to then sell you space on a quantum computer out of their cloud platforms.
Michael Bird: So Craig, to summarise?
Craig Lodzinski: So in summary, quantum computing is completely different to what we know today in classical computing because it uses quantum physics rather than the traditional Newtonian physics that we've been tied to ever since the birth of computers even going all the way back to Charles Babbage. It's not here yet, but it's incredibly exciting as a field. It's absolutely something that there’s a huge amount of weight being put behind from research institutions, from organisations and from governments, and when it does arrive it's going to fundamentally change how we deal with computational problems, not in terms of getting email to people faster or building better CRM systems, but solving problems in the mathematical, chemical, physical space that we simply haven't been able to even attempt beforehand. So I think it's going to be a fascinating leap in terms of how computing is done, it's going to be a paradigm shift in the way we deal with the world, fundamentally, not just in the IT space. In terms of the multiplier effect that quantum computing is going to have further out on every industry and that's going to lead to some very interesting effects that we don't really know yet because there's so many leaps to be done before we get to general purpose, large-scale commercial quantum computing, but the distance in between and the research that's already being done is going to have some fascinating impacts on everybody's lives.
Michael Bird: Craig and Martin, thank you so much for your time, it’s been genuinely interesting talking to you both and learning about quantum computing and quantum mechanics - didn't really think I was going to be learning about that today! Listeners if there's anything in this show that has piqued your interest, or if you'd like to speak to someone at Softcat about anything that we talked about in this episode do make sure you check out the show notes and we're going to put some other stuff there as well. Do also make sure you click subscribe wherever you get your podcast and we’ll deliver the next episode to your device as soon as it lands. So thank you for listening to Explain IT from Softcat and goodbye.
Episode 1: 2019 Tech Predictions
Episode 2: AI and Machine Learning
Episode 3: The Future of IT in Healthcare
Episode 4: Security Trends
Episode 5: 5G
Episode 6: Supply Chain Attacks
Episode 7: Rise of the machines
Episode 8: Unstructured Data
Episode 9: Quantum Computing
Episode 10: Multi-cloud
Episode 11: The Future of Work and Workplace