S2E4: The Future of Quantum Computing with Dr Oliver Brown

What can quantum computers really do for us? Take a listen to Season 2, Episode 4 of insideQuantum to find out!

This week, Dr Oliver Brown takes us on a journey through the hype surrounding quantum computing and find out just what they can and can’t do - and why they’ll never entirely replace classical computers.

Dr Brown obtained his PhD from Heriot-Watt University, and is now a Chancellor’s Fellow at the Edinburgh Parallel Computing Centre (EPCC) and the University of Edinburgh. (At the time of recording, Dr Brown was a Quantum Software Architect at EPCC, but has since been awarded the prestigious Chancellor’s Fellowship.)

Audio editing for this episode by Jonáš Fuksa.

Transcript

🟢 Steven Thomson (00:06): Hi there and welcome to insideQuantum, the podcast telling the human stories behind the latest developments in quantum technologies. I’m Dr. Steven Thomson, and as usual, I’ll be your host for this episode.

(00:17): In many of our previous episodes, we’ve talked about various different aspects of quantum computing, from the hardware to the software and the algorithms to the applications. But where do quantum computers fit in the big picture? What can they do for us and when might they be able to do it? Today’s guest is an expert on high performance computing and might be able to shed some light on these questions for us. It’s a pleasure to be joined today by Dr. Oliver Brown, a quantum software architect and head of the Quantum Applications Group at the Edinburgh Parallel Computing Center, the EPCC, part of the University of Edinburgh. Hi Oliver, and thanks for joining us.

🟣 Oliver Brown (00:52): Hello, and thank you for having me. It’s an honor and a real pleasure to be here and I like “HPC expert”, that’s nice.

🟢 Steven Thomson (01:01): So before we get into the details of what, if anything, quantum computers can actually do for us, let’s first talk a little bit about your journey to this point, and let’s start right back at the very beginning. You’re not working on quantum physics at the moment, but once upon a time you were. What first got you into quantum physics?

🟣 Oliver Brown (01:21): Well, so actually it was really during my undergrad…so I did my undergrad at Heriot-Watt in the physics department there, and we had a nice course, I think it was taught with Patrick Öhberg, about intro to quantum mechanics. I hated all the integral stuff, so that kind of early quantum mechanics stuff. But once we got into the second quantization, things sort of picked up for me and it was really that that I enjoyed. I think looking back with perfect hindsight, yeah, I probably should have gone down the route of doing a computer science undergraduate degree because it’s really that kind of abstract space where there’s none of the messy details that I enjoy a lot more. I like doing matrix vector calculations more than integrals, but I saw a way back to that kind of thing. Just really liked the kind of cleanness of doing second quantized quantum mechanics and really went from there.

(02:13): I started doing a PhD initially with my Master’s supervisor, Sabrina Maniscalco, before she moved to Turku in Finland. And my master’s dissertation with her was “A Serious Game for Quantum Research”. So actually they developed with the computer science department a computer game where it visualized information flowback - sort of the phenomenon of information being returned from the environment to a qubit via non-Markovian dynamics. And then the players were tasked with identifying the regions of the qubit where this phenomenon occurred most strongly, simply by basically a qubit that would expand outwards and back inwards and contract again inwards. And my Master’s project task was basically to group together all the places that they’d identified and check whether they’d done a good job by calculating the overlap with the actual proper states that were most robust in a given environment. And so from that, that was kind of the foundation of my initial PhD project. But then unfortunately Sabrina moved to Turku…well, unfortunately for me, but fortunately for her, she moved to Turku in Finland and I moved to working with Michael Hartman at Heriot-Watt instead. And that project was around stationary states of driven-dissipative many body quantum systems via matrix product operators, which will mean a lot to a niche sector of your audience and nothing to the rest, I imagine. But basically, yeah, it was using tensor network methods to find stationary states of driven-dissipative systems.

🟢 Steven Thomson (03:46): So this was a way of connecting the quantum physics back into the slightly more clean computational side of things that you enjoy.

🟣 Oliver Brown (03:53): Indeed. Yeah, absolutely. So what I actually did in my PhD was quantum physics, but was mostly programming. I wrote sort of 4,000 lines of MATLAB, including the test suite. And during my PhD, I have to say, I quite aggressively targeted EPCC in the sense that I went and took a lot of their training courses available through the ARCHER2 training program because I’d figured out that what I really enjoyed was getting a computer to do all the hard work for me rather than doing all the maths myself. I would suffer it for as long as it took to put together my appendix, that had the derivation in and then I really just wanted to go away and write some nicely engineered software. And yeah, it worked out for me.

🟢 Steven Thomson (04:37): Normally at this point then I would ask about why did you make quantum physics a career? But in your case you tried very hard to avoid it, I guess. So after your PhD then you decided to go into high performance computing and software. Can you tell us a bit more about how that happened and what your job actually involves?

🟣 Oliver Brown (04:58): Yeah, absolutely. So as I say, a lot of it was figuring out during my PhD that what I was really enjoying was that sort of engineering process of creating a software to solve a particular scientific challenge. I think it was clear to me as well that scientific software was what I was most interested in. I didn’t decide I really wanted to be a database engineer or a web developer or anything. It was…I wanted to keep doing science, but definitely on the kind of computing and software side. And part of that as well was I was running the research group’s compute systems. As often happens with PhD students in more computational inclined groups, someone buys a workstation and it’s your job to admin it and do all that kind of stuff. And I knew about EPCC really through probably the…so like you Steven, I was in the Condensed Matter CDT and probably through that and the connection to Edinburgh, I had heard about EPCC and I saw that they had courses that were available for everyone to go along and do.

(05:58): So I did those, learning about Open MP and MPI, none of which I have to say directly applied to my PhD work, which was in MATLAB, but it was nevertheless really interesting to learn about. And I actually started doing Project Euler challenges, which are online just a set of basically numerical problems that you can have a bash at. And I started sort of writing solutions to those in C and with Open MP to try and speed them up, learning bit about parallelism in that way. And then, as I say, I knew I wanted to go to EPCC because that was really interested in developing scientific software. And that’s exactly what - or it’s one of the many things, but a big part of one of the many things I do here at EPCC. And actually I ended up applying for the specific job I applied for because I got the top tip from someone who played badminton with another one of the Condensed Matter CDT students.

(06:50): So there’s a lesson in here about the importance of networking. He worked at EPCC and he was able to tell me which jobs to put a watch for on the University of Edinburgh vacancy system. So I knew I was looking for an application developer role here at EPCC. And sure enough, one happened to come up, coincidentally around Christmas time in the final year of my PhD, which was very handy, and I was fortunate enough to get that position. So I handed in my first draft of my…well not the first draft, but the first completed version of your thesis, handed that in on, something like the Thursday, and then four days later on the Monday on the 12th of March, I started work at EPCC. So it all kind of worked out exactly as I’d hoped it would. But funnily enough, the role I actually applied for EPCC initially was an “applications developer (data science)”. Now, they saw through that pretty much immediately. So my first project at EPCC was a High Performance Computing research project. I was doing some work on a European grant called INTERTWinE, looking at programming models for exascale computing. And then I basically just continued to do more HPC research projects thereafter and never did any data science and eventually became an applications consultant in HPC and then a bit later on, still an architect as I am now.

🟢 Steven Thomson (08:16): So if you hadn’t gone down the route that you have gone down, what do you think you might be doing instead?

🟣 Oliver Brown (08:23): Well, that’s a tricky one. When I was a small child, I really wanted to be a police officer because I watched Waking the Dead. I think a lot of people are like…well, I wanted to be a police officer, but was…Steven can see my glasses, they’re huge because I’m very blind. So that was never really going to happen. But then I watched Waking the Dead and I wanted to be a forensic scientist only. Only I was extremely bad at chemistry, so that ended up in a bin too. And that’s sort of the story of how I ended up doing physics is it was kind of the last thing left. I didn’t like writing large amounts, so I thought, well, okay, I don’t want to do any of the humanities type things. And I enjoyed maths when it worked. That pushed me towards the physics end of things. As I said before, I think with the benefit of hindsight…So I was kind of a victim of, as many people my age were, of IT in schools having been quite heavily influenced by Microsoft in the early two thousands. So when I did GCSE IT…yeah, I should perhaps clarify for anyone hearing my accent. I was actually born in London and raised in Milton Keynes. So apologies for adopting a somewhat Scottish-ish accent.

(09:45): Yeah, when I did GCSE IT was all how to use Microsoft Office. It was Word. I remember doing something in Front Page for web design. Yeah, Publisher, yeah, how to use Excel and Access as well. That was the other one. Yeah, it was so boring. I hated it, hated everything about it. So there was no way that it could convince me to do computing at A Level. But in retrospect, I think that’s probably the route I should have gone down, because what I liked was playing around with computers back then. And what I like now is playing around with computers. Although I certainly don’t regret the way things have worked out. I’m very happy with what I’m doing now and where things have ended up. But the through line has always been for me computing. And I even found a while ago when I was digging through my mum’s basement…or like a cellar, not a basement, we don’t have those in the UK.

(10:44): So I’m digging through her cellar, and I found this essay I’d written for A level physics, I think actually might have been GCSE, but it was like the idea of the essay was what’s the most important or interesting thing in physics? You could write about whatever you wanted. And my essay was a) just unbelievably badly written - I honestly have no idea how I passed because standard of my English was atrocious - and b) was on computational modeling. And my argument had been that if you could model something numerically and computationally, then that was a demonstration that you truly understood it. If you can simulate things, then you actually understand the phenomenon. And if you can’t, then you probably don’t. And I think that is still, I still broadly stand by that statement. I think it’s all very well to say that you think you understand how a natural phenomenon operates, but to really prove that you need to be able to simulate it in some way. Now that may just be purely pen and paper mathematically, right? There’s no reason you have to put it on a computer, but computers sure are a handy way of doing that sort of thing. Now I should say the actual motivation for me choosing that particular topic was in fact, I just really liked playing computer games and it was my excuse to go and investigate physics engines.

(12:01): So in truth, what this is really all about for me is creating a better computer to play video games on.

🟢 Steven Thomson (12:08): I see, I guess that must have been roundabout, what, 2004-5-6, a little while after…was it Havoc that came out with Half-Life 2? And physics engines really kicked off.

🟣 Oliver Brown (12:17): Yes! Yeah, yeah. It was precisely in that era, yeah.

🟢 Steven Thomson (12:21): That makes sense. So normally I would, I’d ask our guests here, what’s the biggest challenge in their research field at the moment? I dunno, does that question make sense with the type of stuff that you do? Is there one big challenge in high performance computing or is it just a case of a new challenge every other week

🟣 Oliver Brown (12:43): For HPC I think the biggest challenge overall is trying to continue scaling up and gaining performance improvements to allow more science to happen. And it is worth emphasizing before I say this next bit, that HPC exists for a good reason. And that reason is that fundamentally simulation a) helps you demonstrate that you truly understand the physical phenomenon, and b) perhaps more importantly, lets you do experiments that you could do, but would be very expensive and time consuming to otherwise do. So, an example of this I always use for outreach actually, is the Vasa. The Vasa is a ship in Stockholm, and if you ever go to Stockholm, I highly recommend the Vasa Museum. So the Vasa was the largest gun ship of its time and on its maiden voyage, it got maybe just outside Stockholm Harbor and sank because it was far too top heavy and just never going to float basically. I mean it floated, but not when they put all the guns on it because they wanted to show off how many guns they could put on a single ship, and it just went straight over and unfortunately took everyone on board with it.

(13:54): But then they eventually raised it and now it’s in the Vasa Museum, and as I say I highly recommend it, but this is a good example of why it’s better to do simulations than it is to build things sometimes. So the point of HPC is to try and save time and effort and energy that you would spend building another more complex experiment by simulating it first. Now with that in mind, it’s important that we try to make these things as energy efficient as possible and justify the sort of sizes of compute that we’re using for a lot of sciences, for example, looking at climate change as well. So it’s important work even with the current kind of energy and climate crisis we may be facing. So it’s important we continue this, but we want to try and make sure that we’re doing it in the most efficient way possible.

(14:40): A lot of that as well is around…computing technology has been in a place where you’ve been able to get continually faster and faster processors - Moore’s Law, Dennard Scaling, blah blah, blah, that comes in every one of these kind of talks. It may not necessarily be fair to say that it’s all over and it’s ending, but it is clear that we can’t rely on getting such gains quite so easily, certainly not order of magnitude gains. I don’t think anyone’s going to come out with a processor next year. I can’t find any wood to touch, but I don’t think anyone’s going to come out with a processor next year, it’s going to be 10 times faster than anything we’ve got right now. We’re not seeing those kind of increases in the hardware. Instead we’re seeing more specialist hardware being added, and that brings us into the heterogeneous computing era. And there the challenges are trying to take that technology that works and make it usable. It’s challenging to efficiently exploit hardware such as GPUs or FPGAs are a good example. It’s a Field Programmable Gate Array, basically a reconfigurable circuit, and it can be extremely efficient, but they’re an absolute nightmare to program because you have to design a circuit. And that’s where a lot of the focus is really is around programming models for heterogeneous systems. And coincidentally also my work.

🟢 Steven Thomson (16:04): Well I guess one of the possible specialized bits of computing that’s coming up in the near future then is the quantum computer, right? People sometimes talk about QPUs or quantum processing units being the next thing that we plug into data centers or supercomputers to give us some new capabilities. So let’s talk a little bit about the quantum computing side of things, and EPCC and you yourself are part of the UK’s new National Quantum Computing Centre and you’re part of the Quantum Software Lab. So can you tell us a bit about what that is about and what’s your role in that project?

🟣 Oliver Brown (16:40): Sure, yeah, absolutely. As you say, we expect quantum computers to ultimately be an accelerator for traditional classical computers. There are many good reasons why you would not expect a quantum computer to simply be a standalone computer unto itself. A lot of the algorithms are in fact hybrid algorithms. And even if they weren’t, you’d still need things like IO and calibration and stuff done by classical computers. So we expect the two to work together in some way overall. And so from EPCC’s point of view, that means it’s another accelerator that we’re excited to understand how it works. But one of the biggest challenges facing quantum computing at this precise moment in time is that the best application that you can point to because it has an algorithmic speed up, is Shor’s algorithm. And you go, okay, look, I can factor these numbers. Now, that’s a fundamentally, I think, deeply uninteresting application. It doesn’t actually benefit anyone to do that, right?

(17:39): It happens that we have used it as the premise of all our classical information security for the last sort of 30 years. So there certainly are ways to utilize something like Shor’s algorithm, but not in such a fashion that it really helps anyone, shall we say. So the big question around quantum computers right now, or one of them is what are we actually going to use it for in a serious way? And coming at it from the point of view of being someone who works in high performance and scientific computing, I’m naturally most interested in how are we going to apply them to scientific problems in to accelerate them. Now conveniently, that’s also an area where we maybe expect to see some of the larger benefits because it’s slightly easier to map quantum problems onto quantum hardware, but ultimately we’re going to have to figure out where these can be used in all sorts of ways.

(18:33): And it’s not just a case of being able to solve a problem on a quantum computer or a quantum accelerator, but actually showing that there’s some benefit to doing so. I could in theory, just completely port any classical code directly over to a quantum computer…it would need to be many times the size of the classical computer and it would be likely zero performance benefit to doing so. In fact it would be far, far worse. But we don’t expect that to be the case everywhere. The question is, what are those applications where really will see benefit? And I think the major focus of my work right now is on applications and really trying to actually use a quantum computer for some kind of benefit and seeing where that line is in various sort of domains and application areas. And the other big thing as I sort of alluded to in the previous chat was that even with classical accelerators, it’s not necessarily easy to make use of them once you have them.

(19:32): So the other big question that occupies my mind is how are we going to connect a quantum computer to HPC? I have two sort of dominant programming models in HPC, which are OpenMP and MPI. So MPI is Message Passing Interface for message passing codes. Codes that need to send data that is privately owned by different processes in between them. And that’s used for scaling things across multiple nodes. So that’s your supercomputer programming model. And an OpenMP is a shared memory threaded programming model. In fact, threaded programming is the more generic title for such things. And that is where fundamentally you have a memory area that every thread of your hardware can see and they just have to do stuff, but they have to do stuff in parallel without stepping on each other’s toes. And often we actually combine these things into MPI+X type programming models where we have some element where it sends messages and some element where it just does threaded programming and that’s all well and good.

(20:30): It’s tough, but it works. And we can introduce GPUs into that environment by having them either talk directly to each other or mediate communications the host, blah, blah, blah, blah, blah. How are we going to do that with quantum computers is a whole other question because for every algorithm in quantum computing, it seems like there’s a completely different interface to the algorithm, by which I mean it’s not going to return a standard data type that looks a particular way. It may return multiple data types. It may be a single shot algorithm that you can just run once and get the answer out. It may be one you to run many times and build up a statistical picture. You might just want the average case returned, you might want every case returned. There’s multiple different ways in which we’ll need to try and interact with a quantum computer and none of it is clear just yet, but we’re going to need to get there because we want quantum computers when they arrive and when we figured out what they useful for, we need them to be usable by domain experts, by people who know the science, not just people like myself who know the computing.

🟢 Steven Thomson (21:34): So this is really the next step that you’re working with then. So we’ve talked in previous episodes to people working on things like error correction and just getting a quantum computer to function, just getting those building blocks to work in a reliable way. And what you’re talking about is once they work in a reliable way, how do you talk to them? How do you plug them into this existing ecosystem and do something with them that’s not just making sure they work? Which right now I think is what a lot of the research and quantum computers is focused on.

🟣 Oliver Brown (22:04): My snappy strap line for this for the group is when, where, and how should we use quantum computing? Where the ‘when’ and ‘where’ is what domains and where do you see the benefit and where do you get the crossover and the ‘how’ is precisely the kind of programming model type questions. How are we going to program them and can we get away from having to design quantum circuits, which is an absolutely miserable way to program anything. Already we struggle with adoption for FPGAs precisely because you have to design a circuit and deal with low level memory type concerns. That situation is not improved on a quantum computer. In fact, it is arguably considerably worse. So I think there’s a lot of work to do there to really integrate them in a meaningful way into the rest of computing.

🟢 Steven Thomson (22:51): I’m reassured to hear you say that because I struggle with quantum circuits. I come from, I guess similar to you, the sort of many body physics background. So analog quantum simulators make a lot of sense to me. The digital ones, yeah, sure I understand them, but there’s a conceptual leap there that I still find quite difficult to make before I can envisage doing anything useful with these machines.

🟣 Oliver Brown (23:12): So I should say as well, for sake of transparency, I don’t really have access to any kind of quantum computing systems right now, but I sort of don’t mind for the kind of work that I’m interested in for the most part. And that’s because what I do have is quite large classical computing and that allows me to simulate quite large quantum computers. And for a lot of things, I think on the software side that’s sort of good enough because if I look at my simulations, that looks like a quantum computer with infinite fidelity of gates, you don’t see that in the actual hardware right now. So at a lot of the time I prefer just to simulate things because then I can check with my algorithm actually does what it’s supposed to before I have to worry about the messy details of hardware precisely for the reasons you just mentioned. But there’s still a lot of very exciting and a lot of rapid progress is happening in these areas, but there’s still a lot of work to do and a long way to go on the hardware side before it looks like something that we actually want to use. And part of the interesting side of my work as well is trying to predict for a given algorithm or application, what are the quantum and classical resources required to scale that to a useful size.

🟢 Steven Thomson (24:24): So in your opinion then, are there any interesting candidate problems where quantum computers are going to be useful and give us a noticeable measurable speed up over classical systems?

🟣 Oliver Brown (24:38): That’s the scary question. So I should say as well, I want to draw a line briefly between a sort of quantum technologies and universal quantum computing. So quantum technologies are a slightly different beast in that you can have very specialist devices that maybe solve one problem very well. And I think in a near term, those are the ones we’re going to see a little bit more of because it’s easier to make those things work. So you can imagine that quantum random number generators, for example, might start to appear for security implement- applications in particular require true random number generation. And that’s a good way to do it. And you can imagine things like the sort of quantum and type devices solve a particular problem and may do it better than we can manage on a classical computer as long as you can pose your problem in such a way that it works. Okay.

🟢 Steven Thomson (25:37): So quantum annealing devices are like the D-wave systems, if I understand correctly?

🟣 Oliver Brown (25:40): Yeah, that sort of thing. Yeah. Basically the issue is you have to set up your problem to look like a Hamiltonian, but then you cool it down and hopefully it just works and the ground state is your answer. So if you can use those, brilliant, you’re away and you don’t need the rest of universal quantum computing necessarily. Provided again that those quantum annealers will scale up to your particular problem sizes, which is not necessarily guaranteed. But I think the applications then for the kind of universal quantum computing stuff where I think there’s still a lot further to go in terms of the hardware before it becomes anything that you could describe as useful. People may point out to me the quantum supremacy paper from Google and things like that, that is a demonstration of something that a classical computer will struggle to do in any reasonable timeframe.

(26:32): But it is not a demonstration of anything that is remotely useful because the circuit that they actually simulated - or didn’t simulate, the circuit that they ran - on the hardware was a circuit that was difficult for a classical computer to run, but not anything else. It didn’t solve any problem. It wasn’t useful for any particular domain or application. Its only purpose was to show something a classical computer couldn’t do. So I want to also draw a line between things that are great demonstrations of the technology, and things that are actually useful are not necessarily the same.

(27:05): So I think the application areas where we’ll see the kind of first useful application of quantum computers is probably places where there is quantum data involved already. So maybe things like quantum chemistry. Now, scale is the issue there, but it’s easier to understand how to map electrons onto qubits, right? Because they are already two-level systems and they’re quantum, so okay. And indeed quantum simulation, so the physicists rejoice - yours is the stuff that’s somewhat easy to map to a quantum computer. The harder things are going to be the more abstract applications and the more general ones. Anything that relies a lot on strong numerics. Again, technology has been making great strides, but IBM are looking at, what, their 433 qubit Osprey next? Now in terms of compute, a qubit is worth a lot more than a bit because you are actually manipulating entangled quantum states, hopefully.

(28:06): So then you’re moving through a much larger state space than the bits provide. However, in terms of input/output, one qubit is worth precisely one bit. So whatever you get out of the end of that calculation, it needs to look like a bit string right of some sort. Now we have a standard way to map floating point numbers to bits. That’s not a problem. The problem is that that standard way requires 64 bits per number or okay, maybe you get away with 32 or even 16. But the point is you need at least that many just to show one floating point number up to a certain precision. And if you want it as good as a classical computer, then 64 and you can’t fit that many of those into a 400 qubit quantum computer. Even fewer if you have to have things like ancilla qubits that aren’t actually used for the result at the end.

(29:02): So there’s a long way to go in terms of the scale of quantum computing before you can easily just map in your kind of classical data in a lot of cases. So instead you have to be very clever about how you map your classical problem onto the quantum computer and more clever people than I are working on those sorts of problems. But it still takes a lot of effort to do that kind of thing. So anything basically the more you need to put that effort in, the less likely it is you’re likely to see a near-term solution, I think is the way to think about it. And also don’t think about anything that requires you to load loads of data into or out of your quantum computer. That one is a no-hoper at any point in time.

🟢 Steven Thomson (29:43): That’s interesting. That makes a lot of sense, but it’s something I never really thought about. So what would you say then is the most important thing that people should understand about quantum computers that doesn’t get talked about in most sort of popular science coverage or media coverage?

🟣 Oliver Brown (30:00): They’re really cool, but universal quantum computers are as of right now, as of today, completely useless to you or I or anyone else. Certain quantum technologies may be useful and quantum computers I think will have useful applications, I’ve not quite staked my career on it, but I’m moving in that direction. I think we will find things to do with them. I’m not sure that I would be happy to guarantee at this stage that they will necessarily do those things much better than a classical computer could. I think that is more of an open question, but I think certainly they will be shown to do useful things going forward. But the big thing and the thing that I sit down…so I do a lot of work with industrial collaborators, and I’ve remembered that part of your question earlier was what is my role in the NQCC software lab.

(30:53): Well, that is precisely looking at helping industry across the UK to sort of understand and interact with quantum computing and what it will mean to them. So really we’re looking at going out and finding use cases for quantum computers within industry and helping people to pilot those and do a little bit of development work on them. And my role actually within that project is I lead the work package around developing those use cases and bringing them in in the first place. So the first thing I do with any kind of new industrial collaborator is I sit them down and give them a talk about quantum computing. And the first of all, I talk, okay, what is a qubit? What is a bit actually to begin with? Because we need to go up from there, introduce as quickly and as simply as I can entanglement, which is not that easy to understand.

(31:42): But fortunately, looking at it as a software developer, I just say, okay, well it’s a state that is not separable, which means we have to have two to the N states and it’s a whole thing. That’s what I need to care about. So do all that. And then I say, okay, and here are the limitations of quantum computing. And I have a set that I always colour in a different colour because those are the ones that are not problems with the engineering, they’re problems with the physics. So that’s things like no cloning, you cannot copy quantum information. So for a start, that means that you would never do many of the things that you use a classical computer for. You would never do on a quantum computer precisely because you can’t copy any data across. So why on earth would you try? It’s clearly going to be inefficient. The data input/output thing where one qubit is worth one bit of information unless you do multiple measurements and slowly build up a picture of the quantum state. As we know, quantum state tomography is itself not an easy thing and scales exponentially.

(32:38): And those are the big two really. But I show them what the limitations are and I say, look, these are the reasons why a) quantum computer is never going to replace a classical computer, and b), I show them the engineering limitations. So the things that are a problem right now, which is things like the number of qubits available, the gate fidelity and the number of errors. And I point out to them that your laptop, my laptop sort of standard commodity hardware does not have error correcting RAM in it because there’s no point because you get errors in the memory of a classical computer so rarely, it’s simply not worth correcting them. Quantum computers are not in that position yet. You get errors a lot, you’ll probably have to rerun your calculation many times to have any confidence in the outcome on actual quantum hardware. And it’s an engineering problem. It’s a problem I believe will be solved with time and effort, but it is not solved today. And the important thing for people to do is to sit tight, begin investigating quantum computing, look at what algorithms and applications may apply in your area, but don’t necessarily stake everything on quantum hardware that’s going to solve all your problems today. That doesn’t exist.

🟢 Steven Thomson (33:49): All right, cool. Two questions to wrap up with then. So there’s one question that I always ask every guest on the podcast. It’s more focused on physics. I guess you’re between physics and computer science these days. But the question is that physics in particular, computer science also has historically been a field for a long time dominated by white cisgender men. It feels like things are starting to change and things are starting to improve albeit far too slowly. So I wanted to ask you, in your experience, have you seen things changing over your career? Have you seen a difference between the kind of computer science side of things and the physics side of things? And as someone who is also involved in hiring decisions these days, how do you approach diversity? How does the EPCC approach ensuring that you have a diverse range of staff?

🟣 Oliver Brown (34:38): So I think I’ve been very fortunate. Well, okay, I’m very fortunate in many ways. Number one of which is I am a white cisgender man and middle class to boot. So really I have everything going for me. And it’s important, first of all to acknowledge and accept that a lot of the luck for me was sort of baked in. But for that reason, I also think that I don’t think my opinion on these things should hold all that much weight for that reason because I haven’t had to face any of the kind of issues that other people have. So I don’t have the experience to know what to do for the best. So for that reason, I think the right thing for me to do and what I’m perfectly content to do is take advice on best practice from people who know better than I do about what the right thing to do is.

(35:27): Particularly I’m thinking of the hiring situation. It is my job to try to be as fair as I can at all times and also to follow the best practice advice that was given me by, for example, the university on these things. And I’ve also been very fortunate. And the other thing I was going to say is that I’ve been very fortunate to have, as I mentioned earlier, my master’s supervisor and original PhD supervisor was Sabrina Maniscalco. And for example, Elham Kashefi - Professor Elham Kashefi - here at Edinburgh leads the NQCC software lab. So I’ve been very fortunate to be sort of surrounded throughout my career by sort of inspirational women in science, which is very lucky for me. But it’s not the case for everyone. And I would say that from what I’ve seen, computer science has very much the same issue as physics. Now, I used to sometimes have arguments about…the thing that gets contentious I think is when you look at positive action type approaches, for example, requiring that 50% of people on the candidate list are representing women and minorities, for example.

(36:44): My view on that has always been that academia is in some kind of equilibrium state that is predominantly middle class white men. And what we know about equilibrium states as physicists is that they don’t change unless you provide some kind of driving action. So if what it takes in order to move that equilibrium to a more equitable state is things like positive action, then I’m all for it. I don’t need any additional luck going my way to help me get a boost up the career ladder. I’ve been promoted something like once every two years since coming to EPCC. So clearly things are working out for me for whatever reason that may be. I don’t need any additional assistance there, I don’t think, but others may do. And I think it’s the right thing to do to make sure that assistance is there if it makes things more equitable. Ultimately.

🟢 Steven Thomson (37:39): I like the equilibrium metaphor actually. That one really appeals to me. I think it’s also, it’s nice to say…we’re both white men. We both have had advantages baked in, as you say. And I always worry that people who make it through this system, they get to a certain point and they think, well, that’s the way it is. That’s the system that worked for me, that got me my position. It doesn’t need to change. Or something along these lines. I always worry the people who make it to the top are…there’s a bit of survivor bias in there, I guess.

🟣 Oliver Brown (38:14): Yeah, absolutely. And I think the same thing applies often to…so for example, I’m in the UCU, I’m in our union, and one thing you sort of notice is that the more senior people get, sometimes the less inclined they are to take part in industrial action. My view is sort of the opposite of that is the more senior I get, the more important I think it is that I do take part. Now admittedly I haven’t hit the really hard point yet where my job is predominantly about getting funding for other people. And I think that is often what pushes people further away from it because they feel like if they’re not doing their job, then they’re not getting funding and they’re not creating opportunities for others. And that is tricky, right? I understand that. But on the other hand, my view is that I’ve been spectacularly lucky to get to the position I have.

(39:04): I’ve worked hard for it too. But it takes…it doesn’t take just hard work. If it just took hard work, everybody’d be a grade eight or above. It takes a bit of luck too. And I’m also lucky because my partner does not work in academia, so I can be a kept man if necessary. So I can afford to go on strike for 18 days if that’s what it takes, but other people can’t. And for that reason, it’s more important for me to do it than perhaps anyone else. And I think the same view applies to diversity and equality. I’ve had all the luck already. If it means that someone else needs a boost, that’s fine. That’s about equity rather than just equality.

🟢 Steven Thomson (39:54): That’s a pretty good way to look at it, I think. All right. One final question to end with. If you could go back in time and give yourself just one piece of advice, what would it be?

🟣 Oliver Brown (40:09): Yeah, that’s a tough one.

(40:15): Probably try to worry less. Things all sort of worked out. I was not necessarily a good undergraduate student, let me put it to you like that. Particularly in exams, that was really bad cause I have no memory. And what really saved me in my undergrad was actually my dissertation. Again, thank you Sabrina. It’s really all her fault that I’m here. And it was because when it came to working on a single project that was kind of my time to shine. I spent the time on it and it all went extremely well and I got great marks and managed to get the grade that I then needed to get into the CDT and blah, blah, blah, blah. In fact, I don’t know if I ever told you this story, but actually Sabrina just forwarded me like an entire email chain that was emails from the central university berating the department for not having offered me an interview yet despite having applied for the PhD months beforehand. And as it turns out, I got interviewed, invited to interview the day - not even the day after, the day - of the internal examiner’s board meeting where they kind of clocked that I was actually going to get the required grade to even possibly be valid to apply or eligible to apply for the CDT. And probably I should have also just worked harder during my undergrad, but I spent a lot of time stressing about a lot of things and now I’m like, you know what? It was all fine in the end.

🟢 Steven Thomson (41:46): Yeah, I think worrying less is a great piece of advice and I wish I knew how to do that too.

🟣 Oliver Brown (41:53): Yeah, indeed.

🟢 Steven Thomson (41:54): All right, so if our audience would like to learn a little bit more about you or the work that you do, is there anywhere that they can find you on the internet, on social media, anything like that?

🟣 Oliver Brown (42:04): Yeah, so I do because I had to find photos of a colleague at a conference in order to prove that he was there for an EU audit. Honestly, that’s why I set it up. I have a Twitter account again now, so I’m @QuHPC there. You can find me starting fights and generally saying things that will probably be career limiting. Occasionally I may actually tweet about work related stuff. There’s also a lot of me being on strike and going to see Edinburgh rugby play.

🟢 Steven Thomson (42:36): Perfect. Well, if anyone wants to check out any of that, we’ll leave a link to your Twitter profile on our own website insidequantum.org. Thank you very much, Dr. Oliver Brown for your time today.

🟣 Oliver Brown (42:47): Thank you for having me.

🟢 Steven Thomson (42:48): Thank you also to the Unitary Fund for supporting this podcast. If you’ve enjoyed today’s episode, please consider liking, sharing and subscribing wherever you like to listen to your podcasts. It really helps to get our guest stories out to as wide an audience as possible. I hope you’ll join us again for our next episode. And until then, this has been insideQuantum. I’ve been Dr. Steven Thomson and thank you very much for listening. Goodbye.