 Hey guys, so we're getting it ready for our next speaker, who is Mark Metakely-Skott. He holds a PhD in information theory and has worked at IBM for more than 30 years, understanding and transporting new technology to customers. He's currently the IBM Q Ambassador Leader, which kind of means that part of his job is this awesome task of spreading excitement about computer computing around, and some might remember him from the previous talk he gave last year about quantum computing in the real world. So let's have a look, sorry, that's the title this year, yeah, I messed up, anyway. Let's head it off to Mark, he's a better speaker than I am. I don't know about that. Thank you that you applauded at the start of my talk, it would be interesting to see if you applauded at the end of it as well. Everything you should never ever do when you talk is apologize, but I'm going to apologize because you should never do it. If I fall asleep in the middle of my talk, it's because I'm incredibly jet lagged. I was falling asleep earlier, but I'm pretty awake now, the adrenaline is doing all the rest. So, let's see, not quite as many people as last year, I have to say, but probably more focused, that's good. So, indeed my name is Mark Metakely-Skott. I'm British, I live in Germany. I would like to take this opportunity to observe one minute's silence for the death of the United Kingdom. Are there any other Brits here, okay? Are there any other, are there any, this is very political, are there any pro Brexit Brits here? Are there any pro Brexit Brits, full stop? Sorry, I have to get that out of my system. I'm so, I'm so unhappy about it. Okay, now let's get back to the topic at hand, which is not political self-destruction, but quantum computing. And, yep, so that's a picture of IBM, one of IBM's quantum computers. We currently have, I think it's 11 or 12 other things online available for you and our customers to use, you in the sense of everybody, if you're not in the IBM Q network, you just want to play with it, there's, I think, six, five or six which you can use. If you, whenever you go to talks about quantum computing, you will, depending on who's doing the talking, you will hear claims about exponential speed up and quantum supremacy, but if you look closely at the technology, you'll realize that there's actually quite a lot of hard work necessary to leverage that to achieve that. We know, for example, we know that there are certain algorithms which really do or really would provide significant speed up, Schor's algorithm, for example, but if you look at the technology, the technology is not there yet. So the question then becomes, what's doable today? What can we actually usefully do today? And the question there becomes, when might we be able to do something commercially meaningful? I mean, that's the question that everybody is thinking about. So I'm not going to give an answer to that. So no, I'm not going to stand here and say quantum advantages in three years or five years or next month, but I hope to throw a little bit of light on how a more differentiated way to think about that. So where have we identified that quantum computing is likely to provide some kind of advantage in the sense of speed up? We know that a quantum computer or qubits as a kind of Hamiltonian solver is, very, very careful choice of words here, is amenable to solving problems of physics from chemistry worse about Hamiltonians anyway. And there's a lot of research, a lot of papers being published, a lot of experiments being completed, among others, IBM with our, with our IBM Q network members on exactly that kind of problem. It's fairly clear that quantum computers are almost certainly going to make a difference in chemistry, material science, that kind of thing, protein folding, exploration, color chemistry. We've already joined the dots there. We've got a fairly good idea that's going to work. The next area where we all think about using quantum computers is in the area of optimization. And that's where it gets a little bit more difficult. It's not like you can say, okay, I have my KSAT general purpose problem and here's an algorithm for solving KSAT on a quantum computer. We're not that far yet. We do have a number of promising approaches to break that problem down and a number of challenges still. But we expect in the area of optimization also that quantum computers will make a difference. Third area is optimal, is third area is in machine learning, actually the fourth area here. For some reason I've got them labeled two and three. In machine learning it's a lot more, I would say today it's a lot more speculative. There are algorithms such as the HHL algorithm. Is anybody not heard of HHL? This is a lecture theater, we're at a university, I expect you to be prepared when you come here. Okay, the HHL, sorry. My sarcastic, formally part of the European Union British humor and combined with jet lag, excuse me if I've offended anybody, it wasn't intentional, so HHL is basically a very interesting algorithm because it provides, in theory at least, exponential speed up to invert a matrix. In practice it doesn't provide you any speed up because it assumes that you've initialized your quantum state with your data and that process currently takes away all the advantage you might get. But it's a building block and we expect that that kind of building block will be integrated in other building blocks to build larger quantum systems. If you look at the application areas, in other words where, when we talk to customers or universities, where are people looking at using quantum computing? This is kind of a non-complete, but non-exhaustive but nevertheless extensive list of application areas and no prizes for noticing that healthcare, life sciences, chemicals and financial services is where most of the action is happening. These tend to be commercial organizations with a significant interest in improving those things. Let's be clear for a bank who are doing risk analysis, a performance improvement of 1% in the accuracy or time that they need to do, particular risk analysis is the difference between being in business and being out of business, so it's a big motivation. So let's have a look at, that's the background, let's have a look at what a quantum computer actually is and to be clear without losing any generality and because we've got some of these things, I'm going to talk about the IBM quantum computer. We heard there are other ways of implementing qubits with some advantages and disadvantages. I'm sure in the next four or five years that's going to really accelerate, we're going to see massive leaps in the capabilities of technologies. We're also going to see one or two technologies basically full by the wayside, but let's take this as an example. So what do we have here? We have a bunch of, basically here, a bunch of signal generators, everybody know what a signal, anybody not know what a signal generator is? You got it, last time I asked a question, hands went up, you all know what a signal generator is. So signal generators basically they generate microwave signals. We have things called arbitrary waveform generators which with those microwaves they make pulses. Those pulses are fed down the plumbing in the chandelier, at the top it's room temperature more or less and at the bottom it's 15 millikelvin. That's an amazing challenge by the way to make that all work. So microwaves pulses are sent down to the chip, stuff happens and then we send other microwave pulses down to the chip to read stuff out, that's how we implement qubits. Like I say there are other ways to do it but actually you're always talking about some kind of control signal and manipulation signal and readout signal. We're using microwaves if you're using embedded ions that might be lasers, whatever, so this is, the next foil is an attempt at an animation, is it working, ooh, is it cheese grommet? Yeah, so here we go, microwave pulses, go down the wires, okay, you get it. Now we need to talk about something we call quantum volume but those microwave pulses and the 15 millikelvin device at the end of the wire, these are all real things, they have real constraints, no signal is perfect, no pulse is perfect, there's lots of anecdotes about how you build a refrigeration unit and how precise and sensitive those copper waveguides are to the extent that you, even the torque you use to tighten a bolt is extremely critical, lots of things that can go wrong, lots of things that can influence performance, so we have a real world device with real errors and it turns out that more or less, I'm simplifying a little bit, but more or less the most critical thing to getting a performant quantum computer, in other words having lots of qubits, having an insanely long coherence time of probably 150 microseconds, the most important thing is the error rate of the individual qubits, if you reduce that means you can play with the other factors, not strictly speaking 100% true but it's more or less how it works, so we measure quantum volume basically by creating a set of random qubit, two gate, two gate operations and measuring the cumulative error rate and that gives us this graph, now, no this isn't graph, this is Moore's law, sorry my apologies, so this is Moore's law and the idea of measuring quantum volume is to find something comparable to Moore's law, so Moore's law was published by Gordon Moore, I think in Popular Electronics 1963, he observed the development of silicon chips, he saw that the die size was increasing every year and that the feature size of the transistors was decreasing, he put the two factors together and said number of transistors is going to double every 18 months, you all know this, so we're always searching for some kind of analogy in quantum computing and the reason we're searching for an analogy in quantum computing is that if you've got a comparable growth rate in some meaningful measure of the power of your quantum computer with the things that are influencing that y-axis where you know what's influencing them, you know what the challenges are, you know what the engineering and physical challenges are, you know what the physics limitations are, so you've got the graph, you can then start to think about oh well, if I get up to so many qubits or this particular quantum volume, then I can solve a particular real world problem, so it gives you more of a handle on how long it's going to take to be able to solve a class of problems, if you look at the idea of quantum volume, then there's a whole bunch of other things which flow into how well you can use quantum computing, one of which is of course the algorithms, so we're all, who's actually working on algorithm development here, can we have more hands, clearly shown please, sir can you hold your hand up, be brave, don't be ashamed, anybody else, he's, he's developing the biggest algorithms, humongous algorithms, so what are you, what are you working on? VQ, okay, yeah, so a lot of work going on developing algorithms, but if we look at the, the IBMQ's systems then there's a whole bunch of other stuff in there which also plays a major role, one that always surprises me is calibration errors, measurement fidelity, gate parallelism, but also things like how good is your compiler, so we, the quiz kit aqua transpiler is open sourced, it is a, a, an active area of both within IBM and in the community of development to improve that transpiler, because just by being clever about how you do gate elimination, how you optimize can make a huge difference, and so if we look at the, the latest machine that we have, which is a 53 qubit machine, that's the topology, then one of the things we discovered was that if you try and entangle too many qubits using Josephson junctions, basically the error rates increase so much that it, it's counterproductive, so you need to, you do that software optimization, and if we look at where IBM is right now on our quantum volume, if you want to definition of what quantum volume is, then there is somewhere there's a link to a paper explaining how it works, we now have four data points, about three months ago we only had three, now we have four, and we've committed to doubling that, so it's a logarithmic scale up here, we've committed to doubling that every year without saying how long we're going to double it, I'll leave that to the senior executives, but it's feasible, it's feasible to double this quantum volume, and then it becomes a, do we have a pointer here, is this a pointer, it's either a pointer or some medical device for doing unspeakable things, just small animals probably, you think I should, like, it's a microphone, I've got all this stuff buried in here somewhere, good thinking Mark, so I guess this, yeah okay, so one of the things you can do, is you can say, well right now with this quantum volume here, here, we can, we can calculate energy behavior of, of battery catalysts, of battery electrodes, or battery electrolytes, when that's a real use case, and there was a paper published a few weeks ago, by Daimler, and ourselves doing exactly that, but then the question becomes, okay what can I do if my quantum volume is here, well, 2024, it's a quantum volume of 1,000 realistic in 2024, I don't know, I won't guarantee it, but it's not infeasible, with a quantum volume of 1,000, or you can do all sorts of, you might be able to do all sorts of interesting things, so what I'm saying is, think about it this way, how long is it going to take to get to where I need to be, if you look at the devices we currently have online, now some of these are publicly available, some of them you can use, some of them you can't, in particular the Rochester devices only for, for organizations in the IBM Q network, we have different topologies, so you can use that to play with different topologies, what you can also do, and I thoroughly recommend people take a look at this, there is a textbook on our API, QuizKit API textbook, where it uses the API, but it's a textbook on quantum computing, it's on GitHub, it's open sourced, and if you want to use it, I think it's, I forget the, I forget the type of license, open commons, no, what's the word, creative commons, so have fun, but we do see, we also see a lot of external committers, or we've enabled a lot of external committers, looking at the API, and the reason I want to do this is because I just want to spend a couple of minutes delving into one part of it, so we, our API, we have these four areas, aqua, which is very much applications, air, which is a special purpose simulator, which simulates the real hardware, it's not an ideal quantum computer, Ignis, which I'm going to take a look at in a second, for doing area mitigation and Tera, which is basically the gate, the gates, gate and circuit level models, so try the IBM quantum experience, it's everywhere you'd expect it to be, so QuizKit Ignis, and this is probably what's going to be of interest to some of you, so Ignis is basically we expose through the Ignis API the open pulse standard, so at this level you are not interacting with the quantum computer anymore with gates or with API calls, you're actually defining and sending microwave pulses to the device, and the publicly available machines, one of those exposes this API, the publicly available one is the only one, and it's a single qubit, don't laugh, it's actually very, still very useful, because it allows you to really look at area mitigation, be my guest, so let's have a quick look at Ignis, yeah, so Ignis allows you to do things like the area mitigation, area calibration, and I've just temporarily forgotten the name of the, it begins with R, the error calibration algorithm, it comes to me in a second, so you can play with the one qubit quantum computer which exposes the open pulse API, and what we often, not always, but what we often see is that developing, what I would say is not ideal algorithms, but things where you think, okay this, this, this is getting towards something that will work in, perhaps in the real world, is yes it's important to know about things like H, the HHL algorithm, or the quantum Fourier transform, or phase estimation, these things are important, you have to understand that, but then you have to really understand the circuit level, you have to understand the, the details of what's going on, and then if you want to take it further and get even more performance, you have to look at this level as well, so all those levels are important, for those of you who are electronics engineers or have some familiarity with electronics, it's a bit like the very early days of transistor based computers, going from architect to literally down to what kind of, what kind of basis semiconductor technology do I use for my chips, how do I interconnect, how do I manage clocks, all that sort of thing, so that's all exposed, this is the mobile machine, which is the largest publicly available machine, it's a 16 qubit machine, and this is Armunk, which is the single qubit, which exposes the open pulse API, and I don't know, has anybody used Armunk, nobody's used it, it's the most, for me the most fascinating part of quantum computing is at this level, because you literally, you're playing with microwave pulses, and the question I often get is, yeah, can you use this API to break the qubits, no, we're not that stupid, we have, we've put, you know, controls and things in there to prevent that happening, either inadvertently or potentially, so it's safe to use, so we are, if you look at what's, where's the road to getting to quantum advantage, in other words, the point where these things actually become commercially useful, and for those of you sitting here, commercially useful, I guess is one aspect, the fun aspect, the intellectual challenge aspect, and of course, is this going to help me some point down the road, is it, somebody going to be sufficiently interested in this, we're at the stage of quantum readiness, or we consider ourselves to be sort of quantum readiness right now, and we anticipate that in the near future, and the near future, so I was born in 1961, which makes me 58 years old, turning 59 in November, IBM will compulsorily retire me in about six years, so the foreseeable future means I expect to see quantum advantage in some applications, areas before I retire, if you belong to, or are associated with, or in charge of, or studying at, any of these organizations, good luck, congratulations, well not all of them, this is where Thomas, look at his face, look at his face, isn't it sad, so not everybody gets full access, for Thomas we had to restrict it, but basically if you're one of these organizations here, so obviously commercial companies, but here, this might be very interesting, or these here, a lot of universities, or these here, so there's currently 102, we broke the 100 at the end of last year, and we've set our aggressive goals to expand that this year, so these are all members of the IBM Q network, and they have more or less access to all the other quantum computers, and if you don't, you aren't a member of one of these, or associated with one, and you don't have access to the IBM Q network, you still have access to all the free machines, and play on those, you may have heard, we announced last October, a collaboration with Fraunhofer in Germany to put together a joint program, there was also an announcement in December that we're doing something similar, but not exactly the same in Japan, and I think that underlines the fact that this is, the whole thing about quantum computing is very strategic, strategic in the sense of strategic importance to economies. I don't know the exact figure, or even though it's about 10 to 12 billion currently being invested in quantum computing, but the projections, all the projections I see is that that's going to increase massively in the next two to three years. Something for you guys and girls here, the quiz kit, quiz kit camps, so there's one due to happen here, taking place in New York, where I've just flown in from, a traverse to Vermont, so March, traverse in New York to Vermont, I guess that's something like those who survive it win, and those who don't freeze in the snow, I don't know, we hold these things regularly, keep an eye out for those, anybody's welcome to apply, we do help and support as well, and that's a good way and an interesting way to contribute and to meet like-minded souls who are obsessed with and in love with quantum computing, like me and you, I hope. Okay, I've, in my jet lagged waffling, I've said enough now, and I think I'm, yeah, I guess questions. One point, one point, okay, here's some some advertisement, does anybody know what this is? Yes, okay, now if I was, if I had the courage to open this, this thing up, which I packed last night, and it's one of the, you know this, when you open the bag up and it goes, that's the state, there's a pi zero in there, so if you go on, if you go on github and search for IBM Q, system one 3D model, you'll find it, so there is a 3D model, if there's anybody with a 3D printer here, please go and print the thing, there's some design files in there, there's also a instruction set on how to compile Qiskit, the Qiskit Terra API onto the pi, and as of two days ago, it now runs on a pi zero, so the idea is we put the pi zero in the model, it will have LEDs, or it has LEDs, and you can actually interact with it, and maybe somebody wants to write the code to actually reflect the, so if you're using the simulator, if you're using the air simulator, you can actually see the internal state of the computer, because you're simulating it, we'll actually reflect that on the model, so an interesting way to interact, and for sure a conversation starter, okay, now. Okay, we have time for two questions, and may I ask to not, like, everyone to leave the room right now, because otherwise it's very hard to understand what people are saying, so are there any questions? One of the key things about quantum computers is the interconnect between the qubits. Yes. Is this factored into the quantum volume as well? Yes. Okay, because if there's not enough connectivity, you're going to basically not be able to use the qubits well. So the thing is, there are qubit technologies where, for example, you have three dimensions, so you have a, try drawing that when you're jet-lagged, so you have maybe a qubit here, and a qubit here, and a qubit here, so you can do, you know, three-dimensional ideal connectivity. With transmon Josephson junctions, you have a 2D lattice. Theoretically, you can do this as well, so the edges, you have connectivity three, and then you have a connectivity of eight. In practice, the eight connectivity is completely useless, because the way you manipulate transmon qubits is with microwave pulses, and it's kind of like every time you fire a pulse at something, it's a quantum system, and every time you fire a pulse at something, something happens, whether you, you know, always parasitic things happening. So that's why in the topologies, the scaled topologies, you see, you know, these large rings of qubits like this. A bunch of qubits. The key question is, what's the penalty when your, when your, your mesh connectivity is less than ideal? The ideal one would be, and this is a, this is a vague attempt at physics irony, the ideal one would be an n-dimensional, n-dimensional space where every qubit was intangible with every other qubit. Here it's not the case. How can you get around that? Well, if your gate fidelity, if your error rates are sufficient, you've got sufficient coherence time, you can swap, you know, you can entangle, for example, these two, but if your circuit wants this one to be entangled with this one, you can actually do that. If you've got sufficient fidelity and sufficient error rates, you can, there are ways to, there were ways to, I wouldn't say fake it, but there are ways to not physically, physically move the qubits, but computationally move them. Then the question becomes, how good does the compiler have to be? And what effect does that have on circuit depth? And the heuristics we see are, the compiler is, a compiler is a compiler is a compiler. The effect on circuit depth is, for most of the algorithms you see out there, like VQE and the stuff people are using, it's, it appears to be polynomial. So I don't know, was that a very long winded answer to your question? Okay, so we're out of time, so let's thank Mark for the talk.