hey everyone welcome to this episode of SEO cyber talk I'm Rick Gupta your host I'm a machine learning research scientist here at the emerging talents Center and with me I have justice and Jason hi how's it going good what's up so we're here to talk today about quantum computing and quantum computing you know I've heard a lot of buzz about it and I'm sure everyone everyone else has too and I know you guys are working on a process really deponent computing so I guess let me start with the first question what is quantum computing right so quantum computing is a new architecture for computing that's being developed and the idea is that we're switching out classical bits for quantum bits and then the question is what's quantum bits in a quantum bit is a very similar to a classical bit but I use quantum mechanics to have an infinite number of states and then once you measure it it collapses down into either a zero one just like in classical bits but this explosion into infinite states gives us a couple of different unique properties that we can leverage so when you say hermit-like you mean like you know I have like a chip and I look at it and it collapses or what's going on there yeah that's always exactly what's happening I mean there's different ways of measuring a quantum bit and observing it but when a quantum mechanical system is working you're not observing it it can be in many different states at once sure and once you actually do look at and measure using different types of techniques it does collapse into one or the other state gosh yes so some measurement is more of an abstract thing that's not necessarily it can be something physically looking at if it could be any way of kind of probing the state of this qubit yes exactly right okay I see and and so so so why bother like what we have great computers already like I can go play crysis that feller has a full resolution like why should I bother you on computing yeah so there's kind to me well the main reason totally would be that gives us speed up in certain situations not every algorithm will see a speed-up when you convert from classical to quantum computers but there are several ones so for instance prime factorization is one where we'll see potentially an exponential speed-up and it has severe implications on cryptography and then the other one is also the amount of space certain algorithms take up so there are certain types of quantum finite automata that take up far fewer bits or qubits then you there classical computer architectures require I see so basically what we're saying is and tell me if I get this wrong right is for some types of algorithms and there's some sorts of problems depending on kind of the structure of the problem quantum computer should give me not only speed up in terms of time but also the space complexity that a classical algorithm would take up compared to a quantum computer yeah that's exactly right okay and so so I'm also assuming that you have to create special quantum algorithms that do this too right it's not just take for example I don't know Dijkstra's and just run down upon computer yeah well for every clone I'm sorry for every classical computation there is a quantum analog that quantum analog is not necessarily faster than oftentimes it can be a little slower but for specific quantum algorithms that we create that you cannot perform on our classical computer you will see these speed ups I see and there is a bit of a merger as to which quantum algorithms might be replicable on any classical computer and so it's actually furthering our understanding of classical computers also see oh that's that seems pretty intuitive to me Jason launched a little bit I I mean I I know you guys have worked on some quantum computing stuff why don't you tell me a little bit about what you're doing and kind of some of the four pieces that we should know about okay yeah so our project is focused on applying qualm computing to software verification of validation more specifically on combinatorial optimization for MP complete or np-hard problems so we're trying to leverage the kind of things that Justice was talking about to attack that problem comic or optimization actually is found in many different applications machine learning and and obviously for software verification of validation so that's what we're working on right now and we're also trying to figure out if near-term quantum computers which are so-called the noisy intermediates well scale quantum computers which I have a small number of cubed and are noisy so there's high error rates if we could actually leverage these Quan computers before we have proper quantum error correction which is what you would need to run Shor's algorithm for example I see and so Shores is the one that s prime factorization correct so basically are you telling you that near-term quantum computing and long-term quantum computing like they're they present fundamental different challenges or yeah yeah so if you think about Shor's algorithm you know you want to break 2048 or si you need that many logical qubits but you actually need like an order of magnitude more error correction qubits so we're really far away from having that number of qubits so you say err correction a lot mm-hmm I mean I I've never heard of air connection error correction at least in our in our classical CPUs or anything what's the deal so so we actually do have air correction in classical CPUs you can implement an error correction scheme where you let's say I have an operational qubit and then you have several redundant qubits to serve as error correction but error correction is not so critical with classical computers we kind of have that under control maybe if you're in some environment which needs it more you'll have it but in quantum computers we have lots of sources of errors that affect qubits so interaction of qubits with each other with the environment and this means that we have high error rates at least right now an error correction utilizes extra qubits in redundancy to perform this error correction and there are schemes that eliminated essentially they can push the error rates down so far that you can do a computation for sort of a extended period of time minutes hours etc I see and so basically I guess what you're saying is that the reason why this is very relevant for near-term quantum computing is that as we have really noisy chips and noisy quantum computing are you saying that we can't do all these fancy things that people have been promising us at 2048 bit RSA you know breaking because we're limited by the amount of error correcting qubits we have yep I see but that's not the only factor but that's one of the major ones I see um yep okay so I mean that's that's interesting because I always thought that you know now that we have a car in a computer right like and a lot of companies do I just thought that oh we can do all the stuff now but I guess that's not true No so so what can we do with your component leaders then yeah so right now there isn't a hole well though we're starting get bigger as our farm computers are starting to push up towards 50 cubits in 100 cubits we're exponentially increasing what we can actually model and what we can do and one of the really low hanging fruit right now is molecular simulation okay so and you mean like like chemical molecules right yeah yeah so so we can accurately represent these these these molecules we can see all the different states and there's a couple of different simulations you can run off of that and as you started getting into only slightly larger molecules such as maybe caffeine it becomes extremely difficult for a classical computer UMass supercomputer to accurately represent these molecules but as we start getting just maybe a couple hundred more qubits this should definitely be accessible again it's it's about the logical qubits versus that keep its need for error correction so actually you know that that brings me to another point is is I always hear about these things that oh you know Google or someone has created a cute a chip that I sent me two cubits or something like that when they say that are they saying that they're sounding two qubits that do actual computation or is that does that include the amount of error correction qubits that they need or is it so what's the deal there so I think the Bristlecone of the 72 qubits there are some number which are there for error correction so I think the Bristlecone has error correction qubits built in most other companies are after using all qubits for logical qubits and they're more interested in finding algorithms which were able to tolerate the noise that is use all qubits for logical and figure out some other way of of dealing with the notes so so okay let's say I have a quantum chip right I can do some quantum computation on it where do I start how do I write an algorithm what do I do mm-hm yeah I mean so you have to you have to go back like justice was saying about designing quantum algorithms which do some computation which has a classical analog and then you have to so there's a software stack in fun computing much like early days of integrated circuit computing and that stack is not well developed yet so you've got a most people have to start the application layer and take that problem all the way down to bare metal so to speak and that process is very difficult right now so there's a lot of development going into exactly how you do that and how you do it optimally so when you're programming a quad computer there's sort of a limited amount of program ability and an application layer that would be accessible for data scientists and engineers but to really utilize them now you have to be sort of a domain expert that has to change yeah they're definitely working towards changing that right now because as Jason was saying every single qualm computer even within the same companies are creating multiple qualm computer chips have completely different architectures there's different connectivity between each of the qubits and this really matters on how you kind of computations you can have and the efficiency of them and each basically company that that creates a qualm computer and then makes it accessible to the public they will abstract a way that compiled a transpiler from normally Python there are there are other programming languages is that they'll use but it's not like Python they'll abstract away a transpiler takes Python code down into a format that they can actually run on their their comm computers which is normally accessible via the cloud so so are you telling me that I could today you know you know galosh my my cloud editor or whatever and just kind of like write a program like I could just use for loops and everything or or like a is there like a specific way after write a program like I've heard of things like quantum circuits and adiabatic or all these things I've heard like what can I just write a program is what I'm asking I mean sure you could but you'd have to understand the way quad circuits up with work so I guess we could say that the abstraction layer which is above quantum circuits at the application layer there's a limited number of applications which are accessible by data scientists and engineers if you actually want a program a corner computer you have to understand at the circuit level and that means understanding of the quantum computation level so is it fair for me to say that if I were to bring quantum computing to an analogous classical setting and if I as a data scientist wanted to use that I would basically writing like ver log for ML is that an equivalent for quantum computing of where we're at now with a quantum sir come on yes so I mean it depends on if someone has implemented the available quantum algorithms that you're looking at if they haven't then you'd have to figure out how to implement it or you'd have to go to the literature and implement yourself I see yeah the whole whole field is in its infancy yeah I would say the future at least near-term –is– should kind of see quant appears as an alternative to GPU where instead of if you want to reach an algorithm that's very difficult to do in the first place on a classical computer right now we would go and use a GPU to give us a speed-up and that worked for most algorithms but then if we even have a better speed up using a common computer you might call that as a subroutine and then get the information back and continue so as far as I know right for from my partner engineering background a little bit is is GPUs are this SIMD architecture which are really good for matrix operations are you saying that quantum leaders can do matrix operations really well or just like are you just using like an analogy that GPUs provide speed up for certain tasks and quantum computers can also serve as this kind of core processor basically so a real accelerator is okay that's a good way of looking so will I be so what are you guys saying that I will never have a quantum computers just like I can't just log into a computer right like it's gonna be like a chip that's kind of sitting on the side that I can use um so what would you mean by logging into the Chrome computer like it's not like I could just like you know come into work and log in on an active directory and a Windows running on computer like it's more like a it saying that it's like a coprocessor an accelerator and you would submit a job to that qpu gotcha it would it would compute I would send back the result and you could read that and you know on a server let's make you do it I see and so in so long term you know what are the goals right so like I understand your term everything's noisy I assume that there's some constraints with like we have to keep it near like absolute zero or like you know really low temperatures what's what are the long-term goals for quantum computing well the one definite long-term goal is to get enough qubits and enough physical qubits so that you can run error correction pushing the coherence times to longer time periods to better run larger problems to do execute Shor's algorithm for example and the sort of timeline that most people feel for that to happen is about 10 years maybe 10 to 20 years so beyond that I think I mean there's a number of alternative quantum computing architectures like topological quantum computing that Microsoft is working on there are other models that there's a whole bunch of stuff in the academic literature that's coming up so it's a lot like in the 1950s when nobody knew that the transistor was going to emerge as the dominant meat medium of computation so I think it's a lot like that so it's hard to say what's going to happen in 10 20 plus years but I think everybody is is aiming at being able to execute yours that's the one thing for sure everyone's aiming for ok sure for sure and so I mean this is this is again very interesting I can't say I understand all of it but if I wanted to go and learn more about that do you guys have any resources for the audience to go read up on this stuff yeah so we're starting to compile on our own quantum hub which will link the link to that what's drawn compiled a corpus of materials that you can look at it will include act deck papers they'll include YouTube videos and books and anything else that we find helpful maybe to some tutorials to get you started I definitely suggest a great way to get intro is to follow some of the main academics in the area especially Seth Lloyd has some interesting stuff on YouTube along with Ronald McDonald well Jason any resources for you or researchers oh geez there's a whole bunch Scott Aaronson Seth Lloyd John Prescott Caltech I forget the guys name at UC Santa Barbara he works at Google any Farhi from MIT who is also Google now I could keep going on ok it's better to put a list yeah we can so so quantum hub is what I heard no that's something that you guys host so we'll put a link to that in the description and on the screen as well and then all these academic papers which we can link to as well mmm that's really helpful again guys if you guys want more information on the work that we're doing you will again link them in the description if you guys have any specific questions that we can answer please you know just email us at info at SEI toxemia edu but Jason justice thank you guys so much it's very informative and hopefully we hear back from the audience really soon thanks for those thank you thank you you you

Faster than quantum computers –> https://phys.org/news/2019-03-artificial-intelligence.html?fbclid=IwAR0I36I9E86-4vRDXtRti0PUVbHxQFLEU3dQhjssAMpEFoQSroFFiMHR-SY

Are you guys providing the link to quantum hub?