IBM's 127 qubit Eagle quantum computer breakdown and reaction to release
![IBM's 127 qubit Eagle quantum computer breakdown and reaction to release IBM's 127 qubit Eagle quantum computer breakdown and reaction to release](/pic/ibm_s_127_qubit_eagle_quantum_computer_breakdown_and_reaction_to_release/a1p4ZE10NzM4VHc_.jpeg)
hello viewers so it seems like every few weeks there's a big quantum announcement it was only a couple of weeks ago that we saw the chinese USTC group release two big results relating to quantum supremacy and so this week what we're seeing is a new announcement by IBM of a completely new quantum computing chip called the eagle it's a 127 qubit chip and this builds upon the previous kind of chips that they've had this is about double the number of qubits that they had before so what we're going to do today is to look at a little bit of the video that they used to announce these results and before that i'm going to walk through some of the basic specifications of this chip so let's have a look at this first so basically this is as i said 127 qubit chip it's in this kind of configuration it's in this so-called heavy hexagonal lattice configuration seeing here is the the colors are the T1 times that's the decoherence time for relaxation from say qubit level zero to one and then you see the the gray or the the colored regions these are of course the CNOT gates where the colors also represent the the fidelity so we'll we'll dive into the fidelities a little bit more but so this qubit configuration is basically what IBM seems to be going towards if you remember some of the other super conducting qubits that people are trying they have this sort of cross configuration this is a little bit reminiscent of their earlier geometries but recently they've been going with this relatively sparse geometry so there's not really many two qubit connections actually the density of the the two qubit connections is actually pretty low but they do this on purpose because basically there's all kinds of issues with crosstalk and spectator qubit errors so at least in their approach this seems to be apparently the best way forward for them so these are some of the T1 and T2 times so this is basically the T1 times broken down per cubit so you can see T1 times around 70 microseconds and T2 times this is the dephasing times this is around about 100 microseconds this is pretty similar to what they had before and so for example if you look at their previous generation machine it's basically comparable T2 times a little bit improved compared to this one but basically it's about the same values and if you compare it to what is happening in the rest of the superconducting qubit world so this is the evolution of the various T1 and T2 times for various types of superconducting qubit technologies and you can see it's kind of exponentially increasing so people are saying this is like a moore's law for superconducting qubit coherence but you know 70 microseconds is more or less kind of in the middle of the range here as you can see in this figure so just in case you're interested the the USTC machine has a coherence time of 30 microseconds so it's a little bit less and actually sycamore has even uh lower coherence times this is the value that i found for a paper it's about 15 or 19 microseconds this is the the readout errors actually so you know you have a particular state and then how well you can actually perform the measurement so average readout error is about uh 0.09 it's about like 10% error in in the readout so actually surprisingly high in terms of readout you know compared to gate fidelities and things like that these two bottom graphs here i think what this is talking about is like the probability of measuring zero when you prepare one so if you prepare one and you get zero obviously that's an error right so what's the error of that you can see it's mostly you know in the sort of ten percent range although there's a few spikes around even like 80 which i assume it means that you prepare one and most of the time it gives zero which is kind of interesting i don't know actually why the measurement errors are so high so the same one thing i wanted to ask why isn't it symmetric based why is there why are the errors not symmetric based on what you prepare yeah i guess because so i'm not really superconducting qubit expert but you know uh there is an energy asymmetry with these qubits right so you know one energy level is actually physically higher than the other energy level so it's not a really completely symmetric system so if they were degenerate yes you would expect it to be symmetric so these are the actual gate fidelities so I think these fidelities are so even though in the web page it's quoted as an error it says one right so i think that fidelities so basically x error square root x they're basically perfect so I think people no longer really even talk about quoting single qubit fidelities anymore because they're basically perfect right the CNOT error of course is a little bit more difficult so they have larger errors and this is basically going to be the bottleneck in terms of the actual you know fidelity of the of the gates in this case so the CNOT gate has basically around about the error is 0.02 so about two percent error you compare that with basically what's happening in the rest of the world so you know it's not about two percent error so this is fidelity so in this in this case would be 98 so you can see some other works are achieving better fidelities you know but they're probably optimized for that might be smaller systems too so yeah certainly not not bad for you okay so that's pretty much all the stats that they actually give there's not actually much in terms more information so as Scott Aaronson has said in his blog basically there's there's nothing he's been asked to quote you know yeah like uh so you know what what does he think of the new developments with this IBM and basically there's no information right so so i think what i showed you is pretty much all the hard data they have they had a so-called state of the union address of the quantum computing a couple of days ago so maybe we can extract a little bit more information from that so let's let's have a look at the video oh that looks fancy I want to welcome you all to the 2021 IBM quantum summit computer this is Jay Gambetta expert or what is his expertise he's a quantum expert yeah so he did his phd with Howard Wiseman who's a well-known Australian quantum physicist he's some yes so he's been leading this quantum computing team at IBM for some time now we need to feed and provide more energy okay for our growing populations well at the same time it's a bit debatable where the quantum computing will help that today as Dario announced just a few minutes ago we broke the 100 qubit barrier so yeah so it looks like they are basically trying to get up to a thousand cubits and then and then a million i guess by 2024 or five okay wow using over a thousand qubits and we still plan to do this by the end of 2023 being and i kind of like the fact that these guys seem to be actual researchers so yeah they came across as like super salesman-ey which stuff i think that would be actually turned off committing to scaling our quantum systems each year and then at last year's quantum summit we released our hummingbird processor at 65 qubits by the way we haven't chosen these targets and scaling commute number arbitrarily in fact the roadmap is fully aligned with the development of critical enabling technologies along each step is working there oh no next year yeah yeah you got a double every year now here's a visual of what our Eagle processor looks like it's two chips the josephson junction based super lithium transplants sit on one chip and is attached to a separate interposer chip through bump i can categorically say that this is the most advanced quantum computing chip ever built in fact not only has it been built but our Eagle has landed. yeah well i do feel a little bit like i'm watching a sales video because i guess we are wait are we buying it yeah are we invited to this conference sorry for this presentation well no you you can just sign up for this anybody can sign up for yeah i didn't actually end up signing up but yes but anybody can here's the device map for IBM washington currently deployed and it is an exploratory system that the team is currently busy putting through its paces with calibrations and benchmarks in fact here's a circuit that Jay actually just ran this morning showing entanglement across four qubits but it's only four cubits okay yeah so i mean i you know this is showing like if you get uh the corners this is like making it a cat state i think right um i was expecting this would be like for a lot more qubits but okay I guess they're just testing it and our plan is to have eagle widely available at the end of the year and most importantly we are on track with our hardware now we next plan the scale to 433 qubits with our IBM Osprey processor in 2022. and for the team is this is a good thing about the superconducting qubit technology which is that once you make one unit i mean just more or less have to print more of them so making them is not the hard part it's it's controlling them effectively and you know i mean D-wave already made you know 2 000 qubit machines it's not that nobody's ever made a 2000 qubit quantum computer it's just that you know these ones are sort of better performing i'd say as i said quality is a measure of how good our technology is of implementing quantum we measured this with a metric we introduced to the world in 2017 called quantum volume what is quantum volume oh it's like a combination of how many qubits you have and also like how deep your quantum circuits are and since then we set ourselves the goal of doubling our quantum volume every year so far we've doubled it every year and we currently have a quantum volume of 128. we have succeeded in improving our T1 times dramatically from about 0.1 milliseconds to 0.3 milliseconds we have
tested several research test devices and we're now measuring 0.6 milliseconds closing it on reliably crossing the one millisecond barrier we have also had a breakthrough this year with improved data fidelities here you can see these improvements color coded by device family our falcon r4 devices generally achieve gate errors near 0.5 times to the power of minus 3. so this one is about 0.1 but then the one that they're quoting actually was more like two percent so i'm not i'm not quite sure whether i guess this falcon thing is the the previous generation right so now in addition to the announcement that we've broken the 100 qubit barrier with eagle we have a second major enhancement we're proud to say in recent weeks we've broken the 0.001 error barrier yes so i think that's mainly the the technical aspect that they maybe there's a little bit more here so this would make our third major announcement today oh we're officially moving from falcon r5 from our exploratory system to a core system 2016 when we put the first quantum computer on the cloud we created a simple programming model where circuits could be sent directly to the quantum computer but this architecture only allows the QPU to work at five to ten percent efficiency a GPU works with a CPU and runtime software to fully utilize this power we see our QPUs working exactly the same way to get the most out of quantum processors we need to bring some classical computing close to the QPU in may we release the qiskit runtime in beta a program like VQE which used to take our users 45 days to run can now be done in nine hours this leap forward and performance combined with our 127 qubit Eagle processor means that from this point on no one really needs to use the simulator anymore i don't get this third breakthrough i mean they seem to be saying that way that they can integrate the you know classical optimizations with the quantum so what they have is the kind of software library that abstracts those things for you so you just like define a problem with your circuit and some parameters and you know you can both simulate it or deploy to the quantum device and it all now integrates those classical you know assisted things as well okay okay hybrid things okay but yeah i mean the thing with these you know variational quantum circuits is that they're not necessarily sort of faster actually and probably depends on the problem but they have they have set up that can automate this yes i don't know you know i mean i've just seen like a lot of stuff where like i saw a paper recently where you know they were trying to you know optimize some kind of energy production and you know they ran it on the quantum computer and they said you know it's so much faster and so forth but i mean you know comparing that to like a just a you know genuine kind of optimizer just running on a classical super computer you know i doubt that those whether that would actually still be like that you know beat classical supercomputing yeah methods so well i don't know yeah i guess it's still but but you know to be fair i don't think it would even be given a chance to reach a full potential if we would just manually do it every time we want to try it so it's good that someone is trying to automate this whole thing at least you can try it and then you know see if it's faster if it's not then yeah you don't use it yeah and you know people will have access to it and do research so maybe they will find a way to make it better yes yeah yeah i just worry that when they present it like this it's like everybody that wants to do any kind of computation they will expect it to be faster just by writing it on quantum and it's just most of the case not not true right it's like it should it's quite specific and i don't think you just sort of randomly do it and it's like oh it's faster great okay i actually didn't know that how do you know when a problem will be faster to solve well it's very tricky yeah okay let's think i mean it's basically like saying how do you find a quantum algorithm that's actually useful when i haven't i see it so it's generally quite a tricky thing and you don't find it randomly okay you have complexity classes that work better okay we call this circuit name class of techniques decomposes a large quantum circuit with more qubits and larger gate depth into multiple smaller quantum circuits with fewer qubits and smaller gate depths then it combines the outcomes together in classical post-processing this allows us to simulate much larger systems than ever before this year we demonstrated circuit knitting by simulating the ground state of a water molecule using only five qubits with the specific technique of entanglement forging water molecule i mean it's that great that's the fact that they could do that is also i don't have much to comment on i mean the ground state of the water molecule is i mean it hasn't it's been known for years so free wise nothing really chemistry-wise it's exciting but i mean okay you can say you've simulated it but you know have you solved essentially have you have you solved the problem exactly or is it the mean field type thing which is much easier well it's five qubits so it's probably just you know it's probably the simplest version of solving for the ground state of a water molecule although i'm just curious to see the paper to see yeah what they did but i i know people like Alan Aspuru-Guzik have been working on you know the application of quantum computing to doing these sorts of electronic structure problems and molecules so it's interesting to see how much better they do i'm just curious we've heard of google doing something what has this one actually done the IBM one yes the speaker actually this one has done something what has he done actually oh it hasn't done anything yes you mean like what circuits is it right yes well we haven't seen anything like that yet so this is all we've seen with there's this video uh and then there's the data which i showed you right at the beginning yeah that's it yeah but the birthday by the end of the year you could use it yourselves oh okay so they didn't test it and well apparently they did a four qubit calculation this morning so i think they haven't tested it very much yeah in terms of um a lot of circuits or anything like that but to me that makes more sense than quantum supremacy he actually wants to do something oh i see what you mean yes after 2023 assuming we continue at the same rate we are in the land of quantum advantage we say the land because we don't think quantum advantage will be a fixed point in time or a specific event we see it more as a continuum where applications start off abstract and esoteric and become slowly more useful over time so you'd think that this is already in the quantum computational advantage regime right yeah it should be it should be right that's what he said though right they were in the land of this well he's saying it's continuum when we don't really uh not a predefined moment or anything like that so you know it's a little bit of a different perspective to say that what the Chinese team or even what Google was saying like that we've reached quantum supremacy now yeah we've done it you know maybe they will compete with in a different task using a task and if they do better in some of them then great yeah yeah so perhaps they have a slightly different philosophy there i have a question does it run on emacs we should if it doesn't it's crime. The D-wave computer what is he able to do right now?
Random number generation no no i mean it's a quantum annealer so it basically it solves the Ising model which is equivalent to MAXCUT and stuff like that oh okay but so we have a review on that coming out i'm just advertising that just throwing that out there but anyway we have a review on Ising machines which should be coming out in Nature Reviews Physics sometime soon and this covers all that and basically compares all the different approaches of solving such Ising machine ising models called Ising machines sneak peak is that the quantum ones don't really perform quite as well as classical quite yet yeah but you know its still early days so don't write off quantum yet but yeah okay so what did you guys think i like it you like it you like that man i like it the aesthetic's really cool it's so futuristic have you seen Tron yeah that's sort of sticking out it looks like tron yeah maybe one should look like Dune then any other opinions are you blown away is your mind blown i don't know not really not really i'm positively surprised well i would say you know what has it done yet i don't know you want to see it do something yes yes like if you saw it factor some you know reasonably large number it does look yes exactly use this simulated water molecule they didn't see how they did the simulation that's the thing yeah i mean i could simulate the ground state of the water molecule right now it would take two seconds i don't know well yeah it's a five qubit calculations it doesn't matter what you did with the five qubits you could of course do that on your computer or right oh yeah but it's the principle it's the principle right i guess it shows it shows that they can apply it to something that's industry wants that's right which is but there is one thing you could not do if you do it in two seconds you could not use the high port quantum i thought you were optimistic well I i believe it's it is there really something you know noble to chase it so it would be really cool if you could do regular molecular electronic structure calculations with quantum computer i just don't know how they do that yet because i haven't seen what a quantum version of like the the really sophisticated electronic structure methods even looks like yet we're still developing those for the classical case so it would be very interesting to see how far they get then we'll leave it there and so thanks for watching and we will see you the next video bye
2021-11-24 18:57