 Thank you. As always, there's always inaccuracies. That's my fault. It's it's now 30 years at IBM. So I was born back in a decade where people called us baby boomers and I originally studied applied physics and electronic engineering and also computer science Back in the days when there wasn't computer science one of the very first degrees offered in computer science and I did a PhD in information theory information theory is basically it's group theory number theory a lot of kind of abstract and convoluted mathematics and a little deal that I realized way back then That actually 35 years later, so I did my PhD 35 years ago 35 years later I would have a renaissance based on what I studied and all the stuff I learned back then Because actually with that armed with those tools You pretty you you're pretty well armed to understand or at least to attempt to understand Because nobody really understands it Quantum physics quantum mechanics and quantum computing So I'm gonna talk about IBM's quantum computing platform Including things like quiz kit the IBM quantum experience and hopefully Open some doors make you think about about what IBM's doing in this space and Of course the appeal to everybody To contribute to work with us. So let's see if this works So the I guess the fundamental question is Have we reached and this is this is our IBM quantum cat The quantum cat he's thankfully alive and Not in superposition and certainly not dead Have we reached the limits of classical computation? So let's have a look at Yes, I've been practicing this I can see you're you're impressed This is this this is 30 years Working for IBM practicing with Microsoft. Is anybody from Microsoft here? Microsoft is love hate. Yeah, it's good for some things, but let's see what happens I Can hear the the the intake of breath here like how did he do that? It's quantum. So this is This is this is everybody's contemporary view of quantum computing For those of you who are observant you've seen me before I was sat in a steppe here doing some stuff So I was quickly writing my foils. I just pull these literally about about three minutes old off Twitter I thought that was quite nice. So we have you know, we are we have the picture of the chandelier And it's cold and it's even more cold and down at the bottom. It's really really incredibly cold And then we have some means of mapping some real-world problem onto probably a Hamiltonian probably using the Ising model and nobody understands what the hell it's all about but It seems to work and of course between this I'm British. I'm allowed to be humorous or at least I'm allowed to attempt to be humorous between this level and And reality there was a kind of mapping and some some details we need to understand and if we take a look at Quantum algorithms in general and We've heard some talks today about different approaches How we might actually use quantum computing But fundamentally we've got to Actually understand a lot about how contemporary current algorithms Work what their limitations are And how quantum computing might might be able to improve that or bridge that gap Is there anybody here with a background in theoretical computer science? Okay So I'm Thankfully for the rest of you and thankfully for me because the theoretical computer scientists will be able to pick holes in it I'm not going to go into a complete derivation of the Bqp hierarchy class derivation, but it really in order to understand quantum computing Or the advantages potential advantages of quantum computing You need to understand a little at least a little bit about theoretical computer science and algorithmic performance in particular as we'll see later one concept, which is quite key is the idea of quantum computer providing a Supra polynomial speed up Okay, it might be an exponential speed up. It might Be something else. It might be super exponential, but it's at least super polynomial And we'll see why that's important in a few minutes This is a foil which does not display also not Okay, so If we look at the class of algorithms which We are attempting to to address with quantum computing We have so-called easy problems. Does anybody know how we classify or Name those easy problems anybody got a an idea polynomial You've for those of you for the computer scientists closure is For the rest of you. These are so-called polynomial time algorithms. So they're algorithms which are a Polynomial in the size of the input whatever that means So classically Multiplying two numbers together is polynomial just in order to do that I have I need a number of operations, which is a which is a multiple of the size of the two numbers I'm trying to multiply together We all know we've all heard or we should have all heard of shores algorithm who hasn't heard of shores algorithm A couple of people Okay, shores algorithm is a quantum It's an algorithm which can be run on a universal quantum computer Which provides a super exponential a super polynomial speed up of factoring numbers So multiplying numbers is polynomial Determining if a number is prime does anybody know what that is It's polynomial Factoring a number an arbitrary number into its prime factors is No, it's not polynomial That's the whole that's the whole basis of cryptography or at least most public key cryptography relies on the fact that that is a non super polynomial and Shores algorithm unfortunately can factor in polynomial time There's a there's a kind of a small gotcha with shores algorithm So for let's say realistic numbers realistic Realistic size numbers you almost certainly going to need tens or hundreds of millions of qubits Currently the the best Universal qubits the best we can achieve is less than a hundred So we're a long way away yet, but maybe one day that will become feasible The original use case for quantum computing was to simulate quantum mechanics That's also a quantum easy problem. We somebody earlier talked about how you can map In particular optimization problems using the variational quantum I can solve a onto a Hamiltonian equation Hamiltonian equations are things you can relatively easily Solver at least observe with a quantum computer. So there are some ways to approach specific problems But in general we're looking for hard problems, you know Typically, these are problems in the area of Algebraic and number theoretic problems Also optimization problems The optimization problem one is of one of the most interesting ones and if I look at the if I look at the space of Developers startups people who are actively working on and with and especially with quantum computing and Apis there's a lot of companies in there in the finance space because if you look at hedging or portfolio analysis or economic prediction These are all fundamentally optimization problems So there's this, you know, there's this this idea that when quantum computing when we get to quantum advantage Suddenly problems in the financial space or logistics, which are intractable to the day will become tractable and I'm sure that will happen and a result of that will be complete change in business business models And how these how these branches how these industries work Machine learning I'll talk a little bit about that in a second and of course simulating quantum mechanics for chemistry Which is the area where we expect quantum advantage to occur earliest Just to illustrate this I don't know if anybody's seen this I'm merely must apologize for HDMI and my my this is Okay 26 So the question is the question was okay The IBM blue gene Q now Supercomputer it can this this stuff you can't read it's saying So for a 512 bit number, how long would blue gene need to factor that? And the answer which you can clearly see here is 26,000 years Okay. Oh, it looks better. How long would it need for a thousand twenty four bit number anybody want to hazard a guess? Exactly Take shores algorithm That's the speed up. So that's super polynomial speed up the small qualifier of this is that you need a fault tolerant quantum computer with Around somewhere 10 20 30 hundred million qubits. So we're still not there yet How do we actually program a quantum computer now? There are a variety of approaches, but fundamentally it comes down to mapping in some way your problem Onto an interference pattern Which is in this interference pattern you set it up on your qubits There are two mechanisms for doing that set up. There's a superposition an entanglement So superposition we've seen you know the block sphere where you put a Initialized qubit into a superposition of two states An entanglement is where I make two different qubits into one system one quantum mechanical system and the real challenge is Given a real-world problem for example inverter matrix How can I map that into a bunch of these types of operations? There was a question somebody asked earlier about different types of Hamiltonians. Who was it? Somebody set up here. Yes, you're asking about new Hamiltonians Right now we we have The superposition the Hadamard transform and the C-notch transform those are things you can More or less easily implemented in hardware But obviously there's a lot of people working on I would say higher level Fundamental Hamiltonian operations which would Where they to exist would also result in speed up in algorithms so we take We take a superposition of all the states This is actually this bit is a lie because in order to visualize multiple qubits with entanglement Using the block sphere it quickly becomes a problem that's Only understandable by pangalactic transformational hyper beings I'm allowed to say that here because you're developers and I can assume that some of you at least have read Hitchhiker's Go to the galaxy who hasn't So for the video there is about three people pull their hand up and all the rest have read it So you know what the mice are right the mice and the hype Okay, so only the mice can understand the the hyper potential hyper dimensional Blocks fear mapping There is a way to do it with two Cubits, but it's extremely convoluted and not really very intuitive Basically, we have a way of entangling the qubits and and making out of a for example 20 qubit quantum computer effectively one one consistent quantum system Then the magic happens where we basically collapse the wave function we observe And out pops the result Every quantum computing program and I don't know if that's People really understand that which is why I'm going to emphasize it again Every quantum computing program is actually a circuit. So it's Initializing a bunch of bits performing some preparatory operations on those qubits and then observing There's no feedback Feedback is forbidden So Reversibility and no cloning to principle operation principles of quantum computing and The output of this quantum program is Non-deterministic it's fundamentally non-deterministic that means When you perform your experiment two times three times four times five times you never guarantee you always get the same result if you're lucky and your algorithm is robust in Non-determinism then you'll you get a very very clear Statistical correlation showing you one particular result and then you've assuming your algorithm algorithm is designed Well, you know, okay, that's the answer there are also inherently in all quantum computers because Even though as we saw the quantum computer is very cold and it's in under high vacuum and There's very little electromagnetic radiation impacting on it It's still not perfect. So every operation on the qubit whether it means putting it into superposition or entangling it or even reading it out Also introduces by its very nature because it's an interaction introduces additional errors and then in fact that's one of the major Challenges in actually building real universal quantum computers is to manage and Handle those errors and keep them as low as possible When we talk about the power of a quantum computer, it's not just important to talk about the number of qubits Because you know, what would 200 qubits what benefit would they be if the individual qubit error rate? Completely dominates any calculation. Otherwise the 200 qubits would be useless. I need to have a corresponding error rate which is Correspondingly low in order to be able to actually make use of that. So there's always a trade-off between error rates error rates in in Fundament simple operations in entanglement operations in readout operations the coherence time of a quantum program which also of course determines how many how errors can add up how errors are Combined and the number of qubits At IBM we talk about the quantum volume, which is a measure of number of qubits and the error rate Also if considerable interest is the coherence time because that determines the maximum amount of time you have for your quantum program and coherence times if the order of tens or maybe a hundred of microseconds are Normal when you think about it back in the day when when Deutsch Feynman and Deutsch were thinking about the principle of a quantum computer And that we might need such things. I can vaguely remember it. I was doing my PhD Early 80s and I can remember this stuff about you know the idea of a quantum quantum bits And I've I've not not looked it up Or maybe I have to do that but I seem to remember that way back then the first attempts Their had coherence times were measured in femto seconds So now we're in tens or maybe even hundreds of microseconds. So we're a long way It doesn't sound like a lot, but it's actually significant development So historically quantum Quantum computing was the domain of quantum and say the quantum scientists physicists chemists We're now at the phase where we're quantum ready and we see this IBM our Competitors I don't like the words are in German Mitch Triter The people in the market in the quantum computing market with us We all see a very very big uptake great interest in quantum computing And right now it's the challenge is to understand how to program quantum computers to demonstrate as early as possible quantum advantage and there is significant economic and business interest and impact on doing that and At some point in the hopefully not too different distant future We will be in the phase of quantum advantage So This is a model or a mock-up or actually might be a photo of the one that actually built. I Think it is anybody know this Of course your developers hackers, you know it. So this is Charles Babbage's Computing machine. It's one component of it It was never built because back in 1823 I think or 1833 they did not have the metallurgy metallurgical skills to produce gear wheels of sufficient strength and accuracy it's I Don't think it's a Turing machine, but it's pretty damn close It's no, it's a Turing machine, but it's Anna. It's a Turing machine But I don't think it's a von Neumann architecture. It's but it's very close So it has a CPU has registers. It has the idea of memory a mechanical Von Neumann machine from 1823 so when my I'm English I'm British And I've been living in Germany for 30 years. So one day after Brexit. I Literally one day after breaks I applied for German citizenship. So I'm now German and British So when my Germans German Fellow Germans they all say well Conrad Susie invented the computer and I pull this out and say No, sorry not not gonna know all the breast ideas that come from British people Which which which raises a very pertinent question about what hell Brexit is Here I am in the center of the center of What what was this this would be there? This would be the Wasn't Star Wars the No the star the the the state Death star so Bustles would be the death star with with the EU Commission. I'm very pro Europe by the way Anyway So and this is this is where we are right now. So there are several of these an undisclosed number sitting at a couple of locations I've been locations You may have seen in the press at CES a couple of weeks ago We announced the first I would say commercially available Quarantine computing system so That's significant because that means we are moving from a pure research Model into a model where we actually going to work with we work with customers on Developing applications And we use for our for our cubits we use These babies here superconducting Joseph some junctions. There are several other approaches Some of which are actually in current use some of which are Currently only of theoretical interest because we haven't no one's actually observed them Specifically the entanglion approach if anybody's not heard of entanglions go Google it afterwards It's entanglions If we actually show that they exist and are able to use them they will represent a very very interesting technology For the basis for quantum computing the key point here is that the underlying technology the methods of mapping Quantum programs onto a specific technology and the method of mapping applications Onto quantum programs these three levels We want to see these decoupled from each other so Such initiatives as open Quasem for the the layer between the quantum assembly language in the hardware open standards for how you actually control and Control and operate a physical quantum computer And also standards for how compilers are built That's all open source. I'll show you in a minute how to get get access to it. There's SDKs and APIs Coming out of your ears on on github but we're So the the important point here to note is this number here Any electrical engineers or electronics engineers here? Yeah So, you know, yeah, Boltzmann's constant every Particularly at microwave frequencies you have equivalent temperature noise floors and all that sort of stuff Before I came up here. I was down at the amateur radio bench down there and Because I'm a radio amateur as well. So talking to the guys about moon bounce and that sort of stuff So this is all This is all very important when you're working at microwave frequencies low noise And that's actually one of them. One of the reasons why the quantum computer has to be so cold Is that you need the noise the environmental noise around the qubits to be lower than the value the equivalent temperature value of the one or zero and There you go. So a hundred micrometers is Is here this device is I guess a few Nanometers Does anybody know if I have a gate physical gate on a chip with the size of let's say Ten nanometers anybody know how many silicon atoms are in there. I'm looking to the electronics. It is Hmm No, it's more it's a it's a it's about it's about 50 or 60 But it's still a sufficiently small number of electrons have atoms that You observe you don't observe your your gate is almost dominated by quantum effects, especially room temperature Which is why Moore's law is basically at least the Scaling of gate size is is breaking down the die size expansion is also kind of hit some limits So that's what one looks like or at least an IBM one if you remember the photos from Rigetti and D wave their quantum computers look astound astoundingly similar There's that's not chance. That's kind of the way you have to do it the interest the thing I always find fascinating was This see these little stripes here, and you can't see it down here. This is sort of Big block of thing here and some stripes here. It goes all the way up to the top So the whole thing is cooled with helium 4 helium 4 is the common isotope of helium Helium 4 has a boiling point of I think somebody probably going to correct me about 3.5 Kelvin There's an isotope of helium called helium 3 Has one neutron less its boiling point is around 2.7 Kelvin You use the difference in a refrigerator to basically freeze down to 15 10 15 milliculb in interestingly the amount of helium 3 on earth is is Rather restricted and it doesn't occur very often spontaneously so there's Which you've put the helium 3 in the fridge you don't let it leak It's expensive So we saw one of these pictures earlier 15 milliculb in down at the bottom a bunch of microwave electronics What actually happens inside there input output modules a whole bunch of control electronics here? What you're seeing there is a photo of one of the research machines obviously in a commercial quantum computer a lot of that's going to be miniaturized and I'm sure at some point in the future you'll have a pizza carton control electronics for your quantum computer for those of you who work in in In compute centers or in in computing delivery, you know the term pizza carton. This is the affectionate term for that the CPUs and memory and disk things you put into a 19-inch racks. They're called pizza cartons So I'm sure that that will be a pizza carton the control electronics at some point and then of course Compilers API's STK's and cloud access and HTTP wrappers and all that sort of stuff and So the idea is that we were going to replace this with this In a certain at us actually in a certain way And I was at where was I I was in Berlin yesterday in Potsdam the Hassel Platten Institute They have a museum of computers there. So some very very old PCs You know so Vax and PDP 11 all this sort of thing and these the really really old computers You won't remember it. I do but To initialize the computer to boot a computer you had to physically enter the the actual first memory locations by hand It's actually very very. It's a very good analogy for how we initialize and set up qubits You have to do it manually But a lot of people working on quantum memory and quantum sensors I'm sure that problem will at some point in the future also if not go away at least be mitigated improved So this is this is the bit where you roll your eyes because It was going so well before now he's not been it So the key thing is I need to get the little. Ah, here we go. There we go. No, I already did this oops. Okay Here we go. So What is how is IBM How is IBM working with the quantum? So there is a a whole bunch of stuff on on github There are various api's but what we've what we said is that Basically a couple of couple of toys and giveaways and there's an app for the For iOS there's even a board game called entanglement, which is actually quite fun And and go undergoing revisions and we've developed a model q keyboard I'm not quite sure what the point is but It's selling like hotcakes If you if you don't like Jupiter notebooks if you are The strange kind of person that doesn't like you. Is anybody here doesn't like using Jupiter. I knew it's going to be you So you're obviously one of these people who insists on using Microsoft yes code. Yes, we've got you covered So If we look at the this separation So quiz kit Terra is basically It's the gate level so in quiz kit Terra. You can do a Hadamard transform You can do a CNOT transform you can do XYZ so you can do the direct manipulations So it's very very low-level quantum computing It's really simple to install pip install quiz kit. That's it. You need an API key There is I'll show you how to do that in a second Then we have something called quiz kit aqua and Quiz kit aqua is interesting because nobody's really addressed that's kind of addressed that but What what would it be like if I had for example in Python an API call which said calculate Calculate binding energy And I give it so brackets. I give it my API key and then I say Hydrogen lithium hydrogen close brackets So where's the quantum computing I? Don't care now. I'm just maybe I'm using the quantum computer as a back-end Maybe I'm using a quantum computer simulator. Maybe I'm using a classical super computer maybe if it's only hydrogen lithium hydrogen I'm using my laptop But if it's a if it's a caffeine molecule, I'm using the quantum computer So for IBM, it's clear to us that the level at which you ultimately use quantum computing You're not good. You're not going to have to take care of There are actually qubits down there or not and that's the the level we work with with with quiz kit aqua With we also have quiz kit air which is also Open source publicly available. This is a For access to and working with a quantum simulator There are There is a open source publicly available quantum simulator out there You can use that there is a IBM has a variant of this that's behind air There's also quiz kit Ignis quiz kit Ignis is Part or at least parts of it are in quiz kit Terra quiz kit Ignis and you get the analogy Ignis is like the You know the magma and hot rocks underneath the Earth's crust So quiz kit Ignis is all about manipulating and working with actual Individual qubits and the technology itself. So it's below the waterline or below the crust of the planet We have a number of publicly available quantum computers Tokyo isn't publicly available. Tokyo is only available to our customers However, Melbourne and the other quantum computers are publicly available. You can go On the quiz kit website you can see the I just did a screenshot about about an hour ago So you can see the current operating frequency So if you remember beforehand the 240 millikelvin the equivalent is the equivalent Noise temperature of a 5 gigahertz signal. We're operating at 4.9 gigahertz you can see the the related timings for For Tokyo so coherence times you can see gate errors and readout errors in We make that all publicly available. We're very open about that Why are we open about that because? That actually determines the performance or it determines whether your quantum computer is Your quantum algorithm your quantum program is actually going to work or not In fact the compiler that sits between quiz kit aqua and quiz kit terror You can tell it. I don't really care which physical machine you use Just give me the one with the slow the shortest queue Or give me the one with the lowest error If you look at the actual though, this is the topology here So this is this basically says that qubit zero one can be entangled with each other But zero and six can't Yeah, they can't so the where there's a connection. There's entanglement possibility Interestingly this one here looks totally different from this one So Melbourne is the kind of it's like a very very linear Tokyo is a mesh Wouldn't it be nice if your at least at the application level the compiler took care of that and said well just give me your Just give me your topology and I will map it to what it's ever is available So we also do that Um go on to github, please It's all open source hack it improve it It's in this is being recorded. So I hope Nobody from IBM sees this but we actually we've had a lot of input from people outside IBM who've made some of the software significantly better Okay, you're all developers. I'm allowed to say that nobody's gonna hoist my petard for that one So you can see up here. There's a hundred and seventy seven repositories in total working with an on and related to quiz kit and Around 2k commits So we have some people out there working very actively with this on developing the software and using it but actually developing the the API itself if you want to if you want to sign up then and And you're a student or faculty member or a PhD student or whatever at any accredited academic institution, please go via on the hub Sign in there If you're not in that category, so if you're a working developer earning real money with a real job Then just go on to the quiz kit org. It's all free We also have some interesting interesting collaboration possibilities for Developers for startups if you want to talk about that just let me know And I would say with that how many minutes have I got left one? Okay, so at that point I have got about another 300 foils Seriously, I do have another 300 foils, but they go into go into things like machine learning optimization in some detail But I think that's enough at least initially Just give you a feel for where IBM is and I'll take some questions. Yes So the question was Yeah, you'd heard of or seen something about a comment that All quantum computing problems will be what was it quantum Okay Yeah, I'm familiar with this this is the the noise beats everything principle Unfortunately, I'm I'm not into that enough to be able to comment on whether on the validity of it or I mean speaking as an engineer. I understand the argument however back in the 1960s when I first took a soldier and iron into my into my hand And I look at the size of integrated circuits back then and the performance and the Intel 4040 which was the first 4-bit Microcomputer on a on a on a chip with a clock frequency of I think 60 kilohertz and the performance figures and noise figures and capabilities and compare that with the couple of billion transistors on this I wouldn't want to bet on noise The noise is always a problem now. You can't get away from it But I think where we are currently with the noise ceiling We're gonna push that way way back So there was another question. Hold on. Hold on here. I wanted to ask about the quantum volume Yes, what's the quantum volume of current chips? Like you're for example, Tokyo. Yes, well, of course Well, the question is he wanted to ask about quantum volume And what's the quantum volume of our current chips like Tokyo the best in the market, of course No, yeah, we think we think we think we are Among the best no, I mean the the what what quantum volume is exactly there is a formula and you could think about you know like qubit equivalent or noise equivalents Ultimately, it's it's the idea is just to have an awareness of it's not just the number of qubits Yeah, you can you could make a chip right now with a with lots and lots of qubits But it'd be useless because noise would dominate you need to do both So hold on before the question here. There was a question back there Can I talk about quiscate aqua and how the chemistry API was implemented? Yeah, that's a three-day intensive course at university Basically Basically, if you're right now if you're developing quantum algorithms you have Four maybe five ways to do it The ones already be mentioned is using oracles So can I decompose my problem into such a simple fundamental building block that I can reproduce it? Scott Aronson the the the father of computational complexity theory would would you know Would be shooting me right now for claiming that The second way is can I Can I adapt my problem to a Hamiltonian of some kind? We already saw that with so D-Wave Using that but regretty everybody's using that using the variational quantum eigen solver take my problem and map it to a Hamiltonian The third way Peter Shaw kind of did it without realizing it is there's this really really really nice nice thing And if you're an electronics engineer or a communications engineer, that's that's the point where you have a leuchttung Enlightenment is that the Fourier transform which is in the real world sucks computationally and the quantum computer is a thing of beauty It's fast and it's simple So if you take The hidden subgroup problem Which says I've got a group of things are the two Two subgroups which are isomorphic which map one to one to each other It's a combinatorial problem basically to answer that you've got to you've got to look at every possible Subgroup so for a group of size and it's two to the end computational steps The quantum Fourier transform you just you just do a Fourier transform of it And you see suck all the isomorphic subgroups appear as Like frequencies when you do a normal Fourier transform that technique you can use that to do use to do phase estimation With phase estimation and arbitrary rotations. You can do the HHL algorithm You can do matrix inversion So you take a Gaussian elimination, which is n cubed and turn it into HHL, which is log n There are a few other techniques Somebody standing up, which means I've got to shut the hell up If you want to talk about that if you want to learn about that if you want to work on that sign up There's a there is a slack channel There's a whole bunch of I be able to Amazon the slack channel. We're very happy to talk to you Have I got to stop now? Yes I'm around today, and I'm around tomorrow. So just hit me if you if you've got questions or you want to know more