Quantum Conversations VII: AWS Bracket and Quantum Sensing

Quantum Conversations VII: AWS Bracket and Quantum Sensing

Recording: Quantum Conversations VII: AWS Bracket and Quantum Sensing

we what we usually do right we do introductions we still can do introductions but let's do them pretty quickly because we need to get to a lot of stuff so first of all let's just make a very quick announcement um um some of you guys already know so i was running quantum uh conversations in partnership with ibm as a community organizer but as a result of this and partnering with my good friends who uh basically went to ibm quantum uh i was invited to join it and so um basically as of monday uh my title i'm an ibmer and my title is um um technical like a system development lead for ibm quantum so essentially my job will be to do uh what i'm doing here but uh i think um uh as a focus with ibm customers and partners and uh but obviously also inviting kind of uh companies and partners into their network right but uh uh so uh for uh because this is a community space and you know we never wanted to be tilted towards any specific vendor or anything right so uh it's uh kind of important for me to make sure that this is run not by me personally right but by a community of organizers so stanford quantum folks uh gracefully stepped up and uh offered to basically help drive it and so uh josh ketuchu who is the current uh event chair is with us here and he will help us run the um the program today and i hope in the future uh uh senator quantum will basically plan talks but also we have other committee organizers i don't think you know uh uh stanford folks uh are kind of you know against that right i i think we can we can basically accommodate uh a set of organizers right so uh i think you know i'd like some quantum folks to become the point of coordination right but you know i'm not going away i'm gonna be helping with this right it's just you know i i you know i shouldn't be driving by myself right so which was always kind of um you know not the case but you know i just want to make sure that folks understand right this is this is a community group um you know it has support from ibm from stanford from others but it's not beholden to any of this and for instance i want to be able always to say here like google is better than ibm in this regard microsoft is doing something cool right and so you know sharp has something interesting i don't think we should be like in this specific space uh you're right there shouldn't be constraints um and um if you want to help organize it please reach out you know i'll help basically you know josh and mert uh who is the emeritus chair uh uh event chair who also you know he offered to to help run it right so i think we really have a good group so that's kind of my um announcement um and uh now let's see we have about 20 people let's do a very quick i would say at this point let's try to do like 20 second uh uh intro so just introduce you say your name where are you from where you're connecting from and what's your relationship to quantum computing uh let's start with you brett uh my name is rick i'm from missouri in the united states and uh i'm just interested in this field in general i think this is uh very early days but uh potentially a wide reading wide reaching ramifications all this so thank you brad uh yoshi next um hi everyone my name is yosh from denver colorado but currently i'm living in california on campus at stanford um and yeah i'm just really intrigued by quantum computing quantum information science uh complexity theory and it's like a potential profession trajectory thank you uh brent lansford yeah so i'm calling in from austin texas i spent seven years of my career in silicon valley working for a stanford startup silicon light machines i'm currently the global chapter chairman of the ieee sensors council and founding officer of the ieee san francisco bay area mems and sensors council i've had a motto since 1999 it's all about quantum so thank you for hosting this fantastic welcome brand uh nick braun hi my name is nick braun i i've been at ibm qana for a long time doing uh experimental concerns i'm coming at you from san juan puerto rico nice great to have you back nick paul lipman yeah hi i'm i'm from the bay area uh accent notwithstanding um i am a cyber security guy i'm ceo of a cyber security company but looking at potentially uh investing in uh or joining a startup in the quantum computing field awesome welcome paul adam liekman hello hello i don't know if you can see me or not okay i'm sorry oh maybe that just did it um i am a local technologist to stanford area and i've recently completed the five course at mitx online with understanding quantum and uh my background's been in vincent's and mems chips and uh packaging and quality systems so then iso 9000 quality systems and been pretty deep into medical and also aerospace great bay area is a fun place to learn new things thank you thanks for hosting this sure cheryl filkus of filicus hi you got it right um i'm an ibm systems engineer in the poughkeepsie office and um i'm uh working on openshift integration on a number of multi-arch systems and um we've been working on z and p systems and i'd love to extend openshift to uh running uh containers that have a quantum capabilities they're fantastic i can't you know wait to meet you internally and talk about it so okay welcome i'll lock you up on blue pages i am already i'm also there uh actually i'm going through like days it's days and days of hr work i didn't even know that's possible but it is oh onboarding takes a month [Laughter] introduce yourself hi uh can you hear me yeah i can hear you yeah um i'm i'm a student i'm at berkeley right now i'm just here because i got the link and i'm interested so that's all very cool that's a perfectly good reason to join us welcome uh jamie gill all right hey everybody um yeah so i'm based out of wisconsin uh currently applying for graduate schools in quantum computing and uh while i'm doing that i decided to take a couple of classes this quarter at stanford university online very cool welcome back jean paul sadia hello i'm uh i'm a student of mathematics i'm an undergrad student uh at a county college over in new jersey so i'm in new jersey right now and um self studying numerical analysis and quantum computing computational complexity very cool welcome i have a train going by so please do not be alarmed by loud noises uh jeremiah coleman great hey everyone um i'm a senior um at stanford university studying engineering physics and economics uh i'm the president of the stanford quantum computing association i'm interested in quantum hardware in general gonna pursue a phd uh and like physics or applied physics excited for the talk thank you jeremiah and the thanks again for uh helping us partner stanford quantum that that's really good and that's a good way to start the year so thanks john candelaria hi um yeah my name is john and i work at stanford university i'm the executive director for the system x affiliates program which is uh a program where uh companies from around the world and and in many different industry sectors fund research at stanford one of the areas that uh you know that we have funded is in this quantum area so i'm uh doing the best i can to keep up with what what the latest uh information and news is let's have you back john uh kunal gosh yeah i hope i am visible okay i'm kunal gosh i'm a business student at stanford graduate school of business i'm currently in stanford california i'm originally from singapore and my background is i was chief technology officer in laminar aviation used to develop optimization based software erp software for airlines and i have a background of delivering software to more than 40 plus airlines awesome it's great to have you matteo messi hey everyone i'm a junior at stanford studying engineering physics uh i'm i'm interested in applications of quantum physics that's why i'm here welcome hotel murrah doesn't come hi everyone uh welcome to the first quantum conversations of the year i'm mad i just graduated from stanford i'm originally from turkey istanbul currently i'm located in new york city um i used to lead the partnerships of stanford quantum computing association but i really like this community so i offered to help alexis with hosting quantum conversations and yeah and this is my housemate santi who's also very interested in quantum computing so he's like uh joining locally with me hi everyone welcome earth and thanks again for helping drive it so you know i think in the future we'll kind of kind of use merge as a point of contact if murder you know you're kill that so kind of because we need to kind of you know we need to uh uh like you and josh probably right on jeremiah so you know if folks want to be like organizers i guess you know email me and i connected with merit and the ocean we can we can get this thing going right like i think the idea is that let's let's get a group of people right and kind of kind of volunteer to to drive us and lead us uh and i think it's really great because merc has been in academia now it has been an industry it's really good to have somebody who is experiencing you know a lot of these things so that's really great thank you america and uh you know looking forward to a good year doing these things and santi welcome too thank you uh oliverick hi my name is olive i'm a global studies major um but i just independently research this and i learn more i think your connection was a little bit flaky but we got the idea welcome uh let's see did they forget anybody i think xi cheng zhang zhang hi uh my name is chong and i'm a cs masters student at stanford i don't know much about quantum computing so i'm just here to learn awesome like the rest of us welcome let's learn together uh and vasily sheif last but not least well yeah hello guys uh i'm a software engineer at crunchbase and i'm just interested in front of computing and wanted to see what progress is being made in this field hello everyone all right and i think we have we got uh i i don't know if we got uh vaseline right yes my name is actually vaseline your gif i am in california i am regular participant in the conversations by the bay thank you for alex for organizing it i i'm also teaching quantum computing for european university and mentoring for the quantum open source foundation um and just a quick plug we just started a new cohort number three are collecting applications for the quantum operations foundation so you can take the and uh look at the application i just shared the link thank you yes that's in the chat guys if you want to check out uh resilience um links it's great to have you all right i think that's that's all we got for now uh so let me introduce our first speaker uh and in case you guys join later that we have a bit of change of plans so we will have brett present uh uh aws quantum talk first instead of the panel uh we'll have one of the panelists mark uh do a talk actually instead and so that will be about sensing right so it's a little bit of a change of plans but we will have still time to have a general discussion uh and i think the general discussion topic uh will still be you know how can we make impact with quantum computing this year because it sounds like a breakout year right but let's let's first uh go into our program so brett coons is our first speaker so brad presented uh at my conference scale by the bay because i used to run a still around the conference in the bay area on classical computing uh but we also already started having quantum talks there and so brett is really great engineer he presented on swift for tensorflow right and and i think it was a really kind of uh good review of uh real advanced technology for software engineers which which which is the topic of uh my kind of uh traditional community right because if you do we do this if we do uh computing at scale we need to like any other area progress from kind of graduate student code to production code and so we're at the point where we need to integrate with a bunch of systems run in production and obviously aws is one of the top platforms for uh kind of scalable uh global computing and so uh it's very interesting to learn what the offer and breadth uh took initiative and and uh mastered that and so he's gonna share it with us uh it's great to have you brad thanks to our global communication come from missouri uh welcome and take it you know away all floor is yours all right um yeah thank you again alexi for having me and then yeah say thank you to everybody else for uh showing up and listening in terms of q and a's what we usually do brett uh we are supposed to type the questions in the chat and you can see them there if and so you can pick the questions you want to answer right away or you can wait and you can at any point invite people to just you know say that question on like my voice as usual right so it's really up to you how you want to drive it uh so you know feel free to pick questions whenever you want and uh folks feel free to type questions in the chat after the end we can have a q a yeah just let me do my flow so to speak and then we can we can do some questions at the end if you all like so yeah thank you all for coming and yeah today we're going to talk about uh quantum dvd in general and then aws bracket a little bit in particular uh at a high level the purpose of my presentation is just sort of to try and give you sort of a burnside view of this field sort of an overview of uh how exactly it works and some different uh approaches that are out there right now so towards that end we'll sort of like try to do a very brief overview of the theory behind quantum computing uh from there then we'll look at the actual sort of hardware and virtual hardware devices are available to run quantum simulations and then as well we'll sort of look at some of the software that exists that you actually use to sort of program these circuits and run things from there we'll look at sort of just some applications of these techniques uh just different fields that i think you know these quantum computing methodologies can be applied to and then at the end we'll talk briefly about sort of this concept of quantum supremacy and the future of this field in general so the basis of quantum computing is quantum mechanics uh i did a talk on this a little over a year and a half ago at a tensorflow world and i'll simply repeat what i said there which is that i would love to throw up one slide and you understand this perfectly but i don't think that's possible but the basis of quantum mechanics would say is this thing that's called the schrodinger equation it can be read quite simply simply being uh e psi equals h psi the e you know already it's the e from e equals m t squared einstein's equation it just means energy psi is a little bit tricky to explain but we might think of it as simply being sort of the her system and then the atrip little hat on it is called h bar this is just a special form of matrix called the hamiltonian and so uh hamiltonians are a way of sort of encoding a state we'll say so the strongest equation at a high level says nothing more that the energy of the system is equal to the state of the system this is quite frankly deceptively simple because a number of interesting observations follow out of this little bitty mathematical equation and as a result it's considered to be the pinnacle of 20th century physics wave functions are an important part of how all this works they come to us by way of classical mechanics where they're used there to describe things like sound sort of interactions of energy and systems but the way that quantum mechanics uses them is slightly different and so i think it bears looking at that as well uh in quantum mechanics we have sort of we'll say the state uh the the the uh wave function itself but then we also have sort of this concept of the conjugate pair sort of the opposite of the state and i think this is really kind of where a lot of the the mind bendy stuff out of quantum mechanics comes because this conjugate literally encapsulates we'll say all the other possibilities in the universe and this is where you kind of get into this sort of like psychological questions and things like that there's a few different ways of working with these wave functions a gentleman named drak a popularized what's called bra and ket notation where we sort of think of the hamiltonian as being sort of a projection inner product space with these two two vectors so sort of the one-way function is going one way and the one going the other way is the sort of opposite or whatever so to speak and then from there we can sort of work with these operators in a sort of linear fashion so loosely we might say that the hamiltonian then is just some real matrix sort of times the sort of the cross product of these two vectors um like i said this is a previous field i saw that you all were talking about the best way to pick it up i don't know that i can necessarily tell you the right way to learn but i thought maybe i could talk about some of the stuff that i studied that i found useful and maybe that will be a value to you the single best subject of single best reference i've found on the subject of quantum mechanics is this set of online lectures it's called quantum physics 130 from the university of california san diego they literally start with uh first principles we'll say and walk you all the way up to deriving the dirac equation for the hydrogen atom analytically which is literally as far as you can do things using analysis i i think you'll find or at least i definitely found that whenever i tried to do all this that i had a lot of holes in my mathematical education and so i had to spend a lot of time just sort of going back and relearning just sort of my uh my mathematical basis for stuff before i could actually sort of make sense of all this other stuff so towards that end you know ideally you could take sort of say a course on linear algebra or something like that but for me i don't like whenever people say you know if you want to learn this first subject you need to go learn another subject and then come back and so for me the way i think the best way to do it is sort of just to try and get a high level understanding of the problem you know as simple of a high level but understanding the problem as possible and then you can sort of add more and more mathematical rigor as desired so towards that end this book called quantum mechanics and simple matrix form a gentleman named thomas jordan is his succinct and concise and clear overview of the subject of quantum mechanics as i have found he does everything just in matrix forms doesn't even use wave functions and then he does everything else which is sort of like triggering trigonometric identities so i think literally somebody would say even just a high school understanding of mathematics and a little bit of how to multiply matrices uh could read this book and sort of understand the sort of high-level math that's going on professor jordan has another book entitled linear operators for quantum mechanics which is just a really succinct high level overview of this field if you're sort of interested in linear operator theory and so i would recommend that as well and then in the process we'll say of trying to relearn all this mathematics i sort of went down the rabbit hole of differential geometry which uh i just you know i kind of wish all physics was taught that way you know just with the source of clean unified house slate so to speak after the fire fight yeah your stream okay so um yeah this this book differential forms of applications of physical sciences is uh just a really uh i i think it's just a lovely book sort of to try and explain physics to the the through differential geometry in particular i became interested in the work of david hesses who spent a lot of time trying to to unify the worlds of differential geometry and quantum mechanics and uh so uh these two these last two books are just basically an overview of analysis they're by this gentleman named alan mcdonald but part of the reason why i recommend his work because he's sort of a a disciple of testis so i i think that's interesting as well and then yeah these uh first few books i've shown you are like dover books on mathematics so for like 10 bucks a pop you can buy yourself a lifetime's worth of math we'll say a quantum competing popper then is simply bringing the uh sort of the concepts of computer science to this uh quantum realm uh there's a gentleman by the name of scott aronson who's done some interesting work in his field but he had this set of lecture notes on the archive that i thought was actually a really concise overview of this subject as possible this set of lectures it's like 10 chapters and it goes wonderfully off in the deep end of quantum black holes and money and some other stuff like that but i thought chapters one and two were just a nice clean overview of these things chapter one is simply sort of reviewing uh the mathematics of linear operators from quantum mechanics and then chapter 2 sort of just breaks down how we sort of bring traditional computer science complexity theory to the quantum realm the basic idea in computer science is that we have sort of sets of what are called np complete problems basically an mp complete problem is just a problem that there's no good way to solve we'll say and by extension the only way to solve it is to test every single possible answer or option so to speak if we would allow to turn a phrase we might say we have to explore the entire universe of possibilities uh i personally sat solvers are one area of mp mp complete problems and this is something i spent a bit of time on in grad school so i that's kind of my window in this whole world and with sat solvers oftentimes to sort of solve a problem basically at a high level you're sort of constructed to say you know is this problem uh valid or not you know is this a real problem and so really what if you look at how it actually programmed at the sat solver level it's actually just all like a gigantic truth statement you know is the statement true or false but in order to know that you have to actually solve the problem by extension then you can come back to that thing the reason i talk about this then is because in quantum computing we have sort of this concept of oracle functions where we can sort of consult a quantum oracle so to speak and it can sort of give us a probabilistic one or zero as to whether or not it thinks that we have the answer and so then for certain problems if we set up our systems correctly what the quantum computer can do is basically explore all these solutions in sub-linear time you can actually go through all the options at a less than o of n complexity and so ultimately this opens the door up for a sort of explosion linear versus exponential complexity or rather for certain problems a quantum computer can provably solve them faster than any traditional classical computer so aws bracket is a service that aws finally opened up to the public last fall but basically then you can actually run [Music] quantum circuits on actual uh real world quantum hardware so you can actually just uh you know use these techniques directly they have a handful of different devices on their platform uh we'll sort of walk through them here but you should be aware of these things they have this righty device which has 30 cubits this is a startup that's been sort of at the forefront of trying to build these quantum computers and they've been able to make them larger and larger they have another one it's just an ion trap device it's 11 cubits aws provides two sets of simulators this first sv1 can emulate up to 34 cubits of state space and then they recently announced sort of this tn1 which can emulate up to 50 qubits but your problem has to be certain structured in a certain way so it may not be able to solve all sorts of problems uh and then also the other devices that aws will provide you access to are sort of d waves uh quantum annealers these don't really work with uh entangled entangled quant quantum quantum entanglement but uh there are there are some interesting parts that we can run on these devices the first generation their hardware had about two thousand qubits but then they also recently announced the version two or whatever or version five or whatever but has five thousand uh like five thousand seven hundred of these qubits to work with so i tried to make a few different demos for you all and i kind of ended up making this one just randomly we'll say but actually i think this is probably the most important slide i can show you literally i've just constructed some random circuits we'll say of various uh sizes and then just sort of wrote down how long it took to evaluate each one so literally we're going from a little bit eight qubit system all the way up to 34 cubits so this blue line is really just not really showing us anything more than as our system gets larger and larger it takes more and more time in order to evaluate as we would expect and basically the process is uh you know looks to be like almost like a nice clean straight line the trick though then is hidden over here on the left side of this graph because this chart is in log units so each of these little gray lines is two orders of magnitude more time to run that particular circuit so from this we can see that like my computer can emulate like a 20 cubit system in under a second but as we go up from there it just starts to take longer and longer to do these things by the time we get to 34 33 cubits it takes about 40 minutes to run a simulation i went to 34 this took about 10 hours but technically it was hitting the memory so my computer started swapping i only have 128 gigabytes of ram on my machine but you can see simply that as we increase the complexity uh the amount of time for the computer to simulate it just starts to grow astronomically we might imagine another rung up on this thing you know up to somewhere around say a million seconds so one more run would say take 10 days to run and if we went up one more rung from there we start talking about three years of simulation time and one more run past that and all of us will be dead before the program finishes so the first demo i think you should run on the aws bracket system is simply a demonstration of what's called a bell's inequality or bell's theorem this is just basically one of the classical proofs of the sort of entanglement phenomena and basically rather than doing it theoretically you can simply sort of empirically drive it so there's a demo of this running on the righetti with the aws thing and you can simply fire it up yourself if we can imagine like uh you know small two two two qubit systems entanglement then we can start to think about uh building slightly larger ones and from there i think we can get into the realm of quantum chemistry i think this is one of the more interesting uh short term or one of the more interesting practical applications of all this stuff we have this whole world of physical chemistry where we sort of like derive the properties of various materials just based basically off of how their electron fields work together so we might want to say like simulate hydrogen atoms and then try to measure their energy levels so there's like classical algorithms in the space you know notably like the hartree flock algorithm but the problem with this then is that like hartree fock can't really capture properly a quantum system so then you start introducing sort of these uh adjustments and whatnot in order to uh you know notably like density functional theory to try and make your predictions match the real world but if we have a quantum system basically effectively we can just simply measure the energy of our quantum system and by extension sort of just derive our data empirically rather than having to do the the math behind it so there's this company called pennylane that's sort of a startup in this space uh they have a whole website they have a whole lovely set of demos and sort of uh tutorials on various different concepts in quantum computing and i think you owe it to yourself to sort of look through those things but the demo in particular i would tell you to run is just simply their uh demonstration of you keep you pqe algorithm on h2 uh that would that would be the quickest way to sort of wrap your head around this stuff yeah and then oh sorry and then they sort of have tools to bridge out to pie torch which is a machine learning framework and so if you have the skills of pi torch this can kind of complement some of that uh this vqe algorithm or variational quantum eigensolver as it's called i think is just really interesting because of the ability in the near term to sort of run these quantum circuits and be able to do meaningful science on top so what we're seeing here on the left is just sort of how the vqe algorithm uh was sort of laid out how that how it actually models the circuit for this paper which came out last year uh and then what we're seeing on the right is just sort of uh the results you know sort of measuring the actual uh electron density by extension the the energy levels to me the most potent part of this graph then is simply the difference between sort of this dark red and this light red line uh what is showing is that basically as they made the system bigger i was able to produce better and more accurate results and so by extension i think in the future as we can build larger and larger these quantum circuits uh sort of using them to sort of run these simulations to sort of uh sort of uh match expectations from the last you know century or so of uh traditional science in this field uh it gives us this really interesting ability to have sort of this feedback loop of sort of uh you know reutilizing prior knowledge to build larger and larger systems and circuits the next demo i think you should run is what's called a grover's algorithm this picture is sort of attempting to explain how the the circuits are laid out for this there's a demo of this for the aws system running on the ion trap device and so you can get running yourself pretty easily uh the key thing to take away from this slide is simply right down here the part where it says repeat o of the square root of n times uh this would be an mp complete problem in traditional computing but this quantum circuit here is able to have an exponential speed up and solve it in sub in time and so by extension this is you know this opens the whole door to this whole realm of these sort of potentially quantum speed ups for these sort of problems i think the other thing you'll notice if you run this demo is that these devices are quite noisy and it's going to be a long time before we're say cracking encryption systems or anything like that so from there you know i thought it would be interesting to sort of start to look at some of the d-wave systems because i think what they do is much more practical for sort of solving problems in the here and now the d-wave systems are constructed around what's called this q-u-b-o quantum i have the acronym on another slide here or ising model but basically we sort of uh take our problem we sort of encode it as a graph we'll say and then we sort of reduce the energy of the system and then in theory um at some point we can find sort of an embedding of the graph that uh correlates with the answer that we're looking for so this paper i took these slides from was sort of uh taking various sat solvers and then converting them to run on the d-wave device and then basically they were talking about a lot of the interesting real world sort of numerical stability and things like that that you had to do and actually get these things to work and so then this is another paper that was on the d-wave that i thought was interesting but effectively what they use the d-wave device to factor a large number something on the order of 200 000 uh which is larger than any other quantum device to date or whether whenever this paper come out but basically they use a sort of some abstract math to sort of convert their uh factoring problem into a uh very complicated sort of uh graph reduction sort of thing and then this picture is simply just sort of showing how the adjacency matrix and by extension what the the d-wave system you know sort of the activations of the system uh whenever it's actually sort of run and so then uh the d wave has all these sort of like optimization algorithms that they have that you can run on top uh there's a bunch of these sort of acronyms in this space we have sort of qalo qubo that was quadratic unconstrained by our optimization and then they have various basically we'll say statistical optimization approaches the one that stood out to me then was simply just the traveling salesman problem uh in grad school we had to write a solver for this using simulated annealing and so this just was very much similar to stuff that i had done before but basically they sort of model the set of cities you convert it into a graph problem and then you can sort of run it on this d-wave device so that would be the one i would suggest you to do in order to sort of wrap your head around this whole thing works uh broadly then the d-wave they have a bunch of sort of graph problems and stuff like that if you're if you're familiar with that literature that basically you can run on this device and uh bring bring over if you if you have a problem that sort of fits in that realm already and then yeah i thought also i've mentioned this other paper this is from ibm but it was simply overview of various different uh they basically tried to implement every algorithm they could find on a quantum device i just thought this was really interesting paper for sort of uh showing you a whole bunch of different potential approaches and uh maybe you'll find one in there that that's of interest to you and then um like uh yeah a lot of my background is around the neural networks and so if we have graphs then we can have neural networks we'll say so this is a paper uh convolutional networks where they're sort of constructing something like a uh the cnn will say but in the quantum realm but they sort of are building these on uh the their own operators and then by extension able to sort of uh attempt to categorize input images as just like a basic uh sort of neural network layer and then this was the other sort of paper in that field it's actually from a little bit before that one but basically they were sort of just building like basically a quantum convolutional and then a quantum pooling operator and then they combine these two together to just basically make like a classical fully connected neural network style approach for categorizing in the stitches so there's a nice demo of this one tensorflow has released this tensorflow quantum package which is just sort of allowing you to bridge out to the wider world of tensorflow and so by extension you can get data sets and stuff like that for free and so they have this tensorflow quantum tool which is uh if you're familiar with tensorflow this is just a good set of tutorials and stuff to run through to sort of wrap your head around some of these different concepts and so here at the end i'd like to just talk a little bit about sort of this uh this idea of quantum supremacy uh so this is uh from the paper that google published last year uh their sycamore device which had 53 qubits and they were able to demonstrate solving a problem that a classical computer wouldn't have been able to do to me if we look at this graph it's basically we'll say kind of the uh the opposite of the graph i showed you earlier but if you look over here say at the left side basically as they are building larger and larger qubit systems it becomes harder and harder to measure and get the results you know to make the system go here but they were able to get this 53 qubit system to output the answers that they were looking for and then over here on the left side of this graph we're just seeing like sort of estimates of how long it would take for a traditional computer to to perform these same results and then this is a paper that came to us by way of china a couple months ago but these chinese researchers were able to demonstrate what's called a gaussian boson sampling this is kind of like a we'll say a large scale physics experiment where they generate a bunch of photons using lasers they do some tricky to sort of get them bouncing around together in sync and then by measuring the results on the other side they've in theory solved a problem that would be impossible to simulate on a traditional computer system so they they claim that they well started like their system will do 43 sort of the the simulation level is similar to that of a 43 qubit system without much difficulty and then they uh but they had better and better runs we'll say and so they demonstrated that their system had uh done this experiment at sort of the 76 cubit level as well so this is the second uh proof of quantum supremacy so to speak that we have now and so i think you know we're seeing the dawn perhaps of those probabilistic processors potentially opening up new doors and new possibilities and so with that i thought you know i would talk sort of about the frontier we'll say this is a slide i stole from aws presentation last year but i think it really illustrates these concepts as well everything i've shown you today is sort of this zone one over here and uh with this sort of quantum supremacy experiments this red dot we might say that we're starting across the you know the phase change or whatever into this realm where uh these quantum computers can do stuff that we've not been able to do yet um and then we can you know we can argue about where this orange line should be some people think that we can do meaningful uh you know real world problems let's say uh like 400 500 qubit systems but basically what's cool to me is kind of you know we know where we are and we know where we want to be and just in the future as long as we can sort of you know go down into the to the right uh someday hopefully uh you know humanity will be able to use these tools some regularity and make them make them easy to do [Music] so uh to recap uh we've we've sort of talked briefly about the theories of quantum computing how it works under the hood uh we've looked at some different applications that i think will become more and more prevalent in the future along the way i've shown you a set of different demos that you can run uh to to get this stuff working with different various hardware and software approaches in this field and then finally i want to say the future is literally people like you and i uh aws i think it's really cool that amazon is making these tools basically just available to whomever wants to run them i think this is just opens up the possibility of you know literally anybody in the world being able to to play with these tools and so i think that's just kind of really cool in general um the um yeah the you know yeah i want to say one thank you to aws i got some they gave me some credits uh to run all this so i i want to say thank you to them for that uh but uh all the demos i've shown you today uh were like five dollars we'll say to run on an actual quantum device and so by extension i would argue that we'll say for the price of a pizza you can you know duplicate everything i've shown you here on your own and so by ascension the only difference for any of you to be a theoretical computer quantum computer scientist and being an applied quantum computer scientist is a little bit of your time and so i would hope that you play with these tools yourself as well with that i'll say thanks for listening thank you so much brad i see there are some questions in the chat so feel free to answer these and meanwhile guys you're welcome to um to ask your questions directly where they're able to see the question in the shot uh question yeah brett i'm wondering you mentioned that for the price of a pizza you can have fun like that uh but i'm wondering how much time do you have to spend actually to prepare on the bracket and get everything in there i remember when i was applying to get from righty access to them machine a year ago it was taking a few months until i get into the list and get access to righty because they give me free access and even credits that i didn't use up but i'm wondering how long does it take to go through the process with the aws okay uh basically you can sign up for aws and then you can load up they have various notebooks and stuff it's based around their sagemaker platform and then basically you sort of program your experiment into the notebook we'll say and then you sort of hit a submit button and then it sends it out to be scheduled on the actual device it depends on the device you want they're not always available we'll say but for most of them we'll say every 24 hours you should be able to get your experiment in and then at the end you sort of get like a batch job back if you're familiar sort of doing machine learning workflows with uh preemptable instances that's something i do a lot of and so it's just to me very much that same workflow where you sort of uh you know type all your code in there hit send and then eventually uh you get a result back but yeah like like i said most of the time under 12 hours your experiment will get scheduled and then yeah not all the devices are uh the d-wave ones are usually available uh the ion trap usually has less of a waiting time but then the the getting ones seem to be the most popular right now at least that's what i noticed whenever i played with it okay thank you all the questions hey brad all right we're together um i was just asking on the chat um we're i'm doing some simulated annealing and i was wondering if i can use aws or like amazon gpus because it's really slow at the moment so do you have any recommendations on how to speed up d-waves simulated annealing through some sort of cloud service uh yeah i wouldn't i wouldn't know offhand um i i don't know the speed difference between d-wave server services and the amazon one so yeah i don't know that i can i jump and try to help i have a kind of question back are you trying to run it on the d-wave or you're running it in the cloud somewhere or on your laptop [Music] uh we're running locally at the moment it's kind of okay yeah so i would suggest then you can get very easily within a few hours probably access to the d-wave if you go to their web page and register and i also just just try to run it on their platform because then everything would be perfect there on the leap okay thank you other questions all right if not then i'll hand it over to josh who will uh continue our program from here thank you brad again let's thank speaker thank you for listening um great all right um one quick second um my computer is glitching right now um okay great perfect uh thank you brett again um for that wonderful talk um and quick discussion um so now we're gonna be moving on to mark casabic uh mark is a professor of physics and applied physics at stamford university his research centers on quantum sensors precision tests of gr and advanced my micro microscopy techniques excuse me um and today mark is going to be speaking to us on a talk titled entanglement isn't just for computation frontiers and quantum sensing so thanks mark take it away well it's great to be here um yeah so i have a i have a deck of slides that can expand and contract uh so depending on time and i wanted to tell you uh about some other uh kind of quantum information uh relevant uh topics uh that i'm involved with so let me get my slides up here um this is this should do it uh is that coming through yeah it looks great okay and then this is the thing that makes it easier to read and let's go okay so uh everybody should see a slide that says the quantum principles of distributed entanglement superposition and so i got a couple intro slides here and then i'll i'll drill into some quantum sensing uh examples but i just want to uh start that by reminding ourselves of the the basic uh fundamental physical principles that underlie the operation of uh any quantum 2.0 device and those are uh the principles of entanglement obviously and also uh superposition and um i don't you know i don't i don't really know the sophistication of this audience but it's i'm based on the past discussion seems like uh you all are up to speed on these ideas so i just uh uh wanted to flash up these cartoons from the uh eu quantum flagship website that came out a few years ago when they were just trying to promote some uh you know general uh understanding of these concepts and i i think this is as as as good as you can do uh in a simple cartoon uh describing these ideas uh and you kind of the idea um with with quantum sensing and quantum measurement is metrology i should say uh is that these principles uh when appropriate applied to the right kind of sensors allows you to build sensors that are just better than anything you can do if you're just uh constrained to using uh classical principles i by those by that i i guess i would define it as you know you don't have entanglement uh in superposition as resources and so then you say okay you can build better sensors so what's the point because you know you're always building better sensors if you're a photonics person you're you know you're always making uh you're getting better control of optical fields and and that enables you to do uh you know better forms of metrology uh but there's there's there's something that's really pretty significantly at stake when you when you improve a sensor uh and that is that uh it takes you over a threshold you know from being able to observe something that you previously just could never have observed and and it's it's it's it's truly a threshold effect uh you have to kind of see it in action become a believer in this statement but uh you know there's often thought that well you know if i have a bad sensor i just wait a while and average this output and then i'll be able to see something and if i had a better sensor well maybe i don't have to average as long to see the same blip uh but in fact though the world doesn't work that way you always have a finite amount of time to make your measurement there's always some some background noise that you have to have to defeat and there are a lot of measurement classes where if you just improve your sensor by like a factor of three or four you you go over a threshold between from being able to observe something that that you couldn't have previously observed and that observation can be a discovery i mean it could be an observation of a gravitational wave it could be uh you know discovery of dark matter and you know the fundamental physics side it could also uh be on the technology side of you know a a new form of an observation of us a small protein that you just previously couldn't have observed now you can observe that small protein and without destroying it for example and that can enable like a drug discovery pathway okay so i have a couple of examples that i run run you through quickly uh and and then maybe we can we can talk or you know have discussion uh so uh the bread and butter in my lab is atom interferometry and so i want to begin uh what we do is we build uh devices analogous to interferometers for light and optical interferometer is you know beam of light comes in hits beam splitter you get two separated paths for the optical fields they come back together they uh because they're waves they interfere and the lighting that goes out one chord of the interferometer or the other an amazing thing is when you do this with single photons then that interference uh pattern intensity basically tells you about the probability of a photon clicking in one port uh versus the other port and uh you know just thinking about the the uh you know those those types of experiments really uh you know kind of turn you on to the the mystery of quantum mechanics so uh in my lab uh we do this kind of knock center interferometer stuff with atoms and so uh on this slide i show you a false color image and what you see are two spikes on the left and right hand sides of the screen uh each spike has about ten to the fifth atoms and uh what i want to claim is each atom that is in this these spikes uh this is a montage made from a series of uh pictures in our apparatus i think i got a picture of the apparatus in a few slides each one of the atoms in those spikes is in a in a coherent superposition of those two spatial locations separated by uh 54 centimeters half a meter it's the distance between your elbow and the tip of your your your your pinky for example kind of a macroscopic amount and uh when i say that it's in superposition what i mean is the atom is literally in both places at once and so you say well okay you prove it to me and so the way you prove something like that is you do an experiment where you bring the atoms back together uh without detecting them and see if you can observe something like wave-like interference if you observe wave-like interference after separating for a half a meter then bringing them back together then they they sort of had to be in this this uh superposition state so let's do that experiment uh so you know here's here's uh some process data again the the false color is proportional to the number of atoms in a given location and what i'm showing you is interference fringes at the output port of uh an atom interferometer and what i mean by that is i've i've taken those two half half meter separated uh uh ensembles brought them back together overlapped them and then i actually remarkably observed them to uh interfere and um you know you can marvel that that nature let you do this it's i i think it's it's just fundamentally amazing that um nature will allow such a such phenomenology but you know if you've been thinking quantum computing for a while then you're you know that you're you're no longer you know you know that's just a principle that you can engineer with and and uh it's part of the strange world uh we live in so how do we do that in in in practice well here's here's our an apparatus in the basement of the variant physics building uh and what we do is at the bottom of the pit on the left we have a source of ultra cold atoms we use laser light to uh refrigerate them but no cryogens involved it's just a momentum exchange between beams of lasers and clouds of atoms and if you if you tune the laser frequencies correctly in intensities correctly you can't help but to refrigerate the atoms down to micro kelvin temperatures where their speeds are measured in centimeters per second and then further cooling methods gets them down to uh pico kelvin temperatures where their speeds are measured in millimeters a second and microns per second and uh very different regime from you know the classical world that we live in where atoms fly around in the room at kilometers per second at these low speeds due to kind of the broly's principle wavelength of the particle is inverse to its momentum the associated wavelength is large and and hence that allows us to think of the atoms as waves and manipulate them as waves and it allows them to do the things i showed you on the previous slide like when the waves overlap if the apparatus is you know properly configured they'll interfere with waves um the incident on the right just to give you a feeling for you know what's what's going on uh the in the actual max under interferometer we build that's the the buzzword for the topology we use uh the kind of false color indicates the probability of finding the atom in a given position the yellow arrow shows the distance scale and we coherently manipulate the atomic wave function using pulses of laser light and by appropriately designing the pulses of lasers laser light we can impart momentum to atomic weight packet each photon carries momentum h k h bar k h bar being planck's constant over 2 pi and k being the propagation vector of the light field 2 pi over lambda and that that momentum uh you know interplays with the momentum of the atom mv non-relativistically and so if like the atom absorbs a photon it its momentum has to change by conservation of momentum and just by by engineering the quantum process of absorption and stimulating the mission you can you can make this sort of thing happen with the wave function and uh you know miraculously uh after the atom is separated and come back together uh you see interference fringes uh when the thing is appropriately configured now let's let's just think about physicists love to do these order of magnitude type of uh thought experiments how big is an atom atoms like uh half an ancient to an angstrom and uh let's call that half an angstrom 5 10 to the minus 10 meters um sorry of 5 10 minus 11 meters and and how how big is a half a meter well it's 0.5 meters it's a factor of the size of the atom to the separation scale is is like 10 waters of magnitude it's if you wanted to do an equivalent thought experiment it's like if the atom is a golf ball you've sent half the atom to the moon and brought it back and then you observe uh you know interference fringes now it's just a slightly misleading example because the characteristic scale length for the interference is the de broglie wavelength which is is much larger than the actual size of an atom but you get the point and so part of the uh the technical difficulty in making something like this work is that uh you have to have it nearly perfectly aligned it's sort of like building quantum and quantum gates you know if if anything's missing then that that you do not see in the interference the superposition is not manifest you just see uh it looks like two ensembles of atoms that overlapped and they they they don't interfere this is why it took 10 years to get this experiment to work okay so what's it good for what's the disruptive measurement that that can be done with such an apparatus i'll give you a few examples in the context of the physics department we're using this apparatus to test einstein's equivalence of principle and how are we doing that well turns out that the uh here's some interference fringes on the on the right and a schematic of what's happening with the atoms on the left as they launch up a tube and they hit the optical fields that split up and recombine the wave packets and then they fall back into a detection region we see the interference fringes by the way the previous fringe i showed you previous illustration this one here i took out the sag do the acceleration due to gravity just to show you what's happening with the weight packets so actually these things are all doing parabolas under the they're free falling due to gravity as well as separating and recombining well think about the quantum physics of this situation one-way packet is seeing a different gravitational potential than the other weight packet all right and so then you say well that means it's a different gravitational potential energy and we all know how the phase of a wave packet evolves uh you know potential energy it picks up a factor of potential uh divided by planck's constant times time and so that's the phase and so uh what happens in the gravitational potential is there's a phase shift due to the acceleration due to gravity and that phase shift is huge it's like 10 to the 10 radians difference between the two trajectories of the atoms let's say the lower path of the max under and the upper path now what does that mean well it means that you can make a very precise measurement of the acceleration due to gravity like at the part and ten to ten level g changes by apart and ten to the ten and the fringe shifts but enough that the the atom instead of coming out of one part of the interferometer comes out the other part of the interferometer uh okay how much is 10 to the minus 10 of the acceleration due to gravity well um the mutual acceleration between two people due to gravity two people sitting next to each other in in an auditorium is like 10 to the minus 9g this apparatus is is like sensitive enough to see you know like changes in a person's mass at the you know the level of drinking a cup of coffee you know and that's meant as manifest in the uh the the acceleration due to gravity so it's extremely sensitive uh of a way of observing acceleration due to gravity and what we do we're interested in testing science equivalence principle is to ask do two different isotopes of rubidium which have different nuclear structure different mass compositions they see gravity in different ways and what do you mean what i mean by different ways do they say different gravitational phase shifts and uh like the standard theory says they should see the same phase shift if einstein was right and our our concept of gr is correct the the phase shift for vividium 87 the phase shift for vitamin 85 shouldn't um it shouldn't depend on the mass and so uh we just published paper uh recently where we we showed that at 12 digits uh with with atomic weight packets that in fact you know the einstein equivalence principle is correct that point here is that you know one of the things you build a better sensor for is to uh do some some some fundamental science uh but there's there's more that can be done so right now we're uh here's a picture um from a company which i co-founded called ale sense where we're building practical sensors based on these interference methods uh one thing you might want to do is fly one of these ultra sensitive gravity instruments uh in low earth orbit and measure the gravitational field of the earth um why would you do that well because you know kind of gravitational field that you see in a satellite orbit depends on what's on the surface of the earth like uh and there are certain regions where there for example i you know it depends on the amount of water that's in the region directly underneath the satellite uh and we care a lot about uh what i should say geophysicists do uh hydrology and measuring the earth's water table because uh you know as as as with climate change uh the ice caps mill uh the water in that has to shift where's it going there's not a lot of good ways of measuring that on a global scale and so there's tremendous interest in putting up precise uh gravitational sensors in order to do the simple thing of measuring the height of the earth's water table uh globally uh and so on the right there's a false color map where you you see regions of more and less uh you know water and it turns out it's sort of amazing with these these these geodesy missions uh it's it's possible to we think measure water table heights at the at the millimeter level which is is enough to be uh really geo physically interesting and so uh heck just the other day i got off we were on teleconference with uh nasa goddard space flight and we're progressing in a you know along the path to uh such a mission and so it'd be kind of cool you know if we use quantum interference to actually do something for climate science there's still i don't know we're talking about a flight data 2025 and those probably are that's probably an optimistic uh uh estimate what else well uh coming to smaller scale instrumentation it's possible to build uh really exceptional gyroscopes and gravimeters things that you like just on the on the earth you say i want to measure the acceleration due to gravity let's not separate the wave packets by a half a meter let's separate them by millimeters it turns out you still can make amazing sensors and so uh out of ale says we're building gyroscopes and uh gravimeters and other sensors based on this fundamental quantum interference property which entangles the position of the atom with its internal state uh we're building those uh for applications in navigation uh and geodesy and you know when i when i say you know like you know you when you build a sensor there's certain things that you can and can't do uh and when you go over a threshold well it's you you it's it's really it's sort of uh uh it's it's engaging to look at where those thresholds are for various technology applications and for example in gyroscopy that you know that's the gyroscope is a a key element in an inertial navigation system something that tells you where you are if you don't have gps it's it's kind of amazing what you can do uh as in terms of being able to navigate uh without gps when you have super precise gyroscopes um and i can you know i can drill into some of those applications if you're interested in discussing them uh and then coming back to the basic science another thing that i'm doing with uh my colleague jason hogan jason as principal investigator on this project is configuring a very large-scale atom interferometer out at uh fermi lab in in illinois to detect gravitational waves and to look for dark matter and the principles are similar to uh you know what we do when we look for violations of the equivalence principle and so forth in a nutshell you're looking for anything with changes the position of the wave packet that you know either due to gravity or some exotic exotic force or potential and uh that that change in position is manifest in a shift in the interference fringe and uh we think that by building 100 meter scale instruments we can we can make some new statements about say dark matter uh so that's kind of the what i have to say about superposition and uh you know from the point of view of a quantum computation audience uh you know it's it's what we're doing a single particle with those it's it doesn't involve like creating massively entangled states and and manipulating them and controlling is just these are single atom wave functions but we control them extremely well and uh from a hardware perspective uh you know what a lot of us are seeing who are my company also is involved in quantum computation hardware and so forth what a lot of us are seeing is there's a there's a convergence between the the the kind of the engineering that goes into quantum sensing applications and the engineering that goes into quantum computation application for certain types of quantum computation i mean i'm all atom based uh you know you you need really great date ops you need to do rotations uh almost perfectly and and that's what we need to do when we build those sensors and it's it's also you know what we need to do when we we think about networks and so forth all right turn the page let's let's but let's let's now become you know quantum 2.0 people and say well you know i'd really like to do more than just position internal state entanglement uh how can i use entanglement as a resource to improve you know measurements and so uh here what we do is we we create massively entangled states when i say massively entangled uh some of the work i'll show you we're entangling the wave functions of thousands of atoms and uh these thousand atom uh entangled states are not controllably entangled in the sense of a you know a quantum register uh or you know a lot of quantum registers but they're they're entangled in such a way that uh allows us to uh do metrology um and phase metrology at levels that are beyond anything we can do with with single quantum systems and so uh the states we make are essentially uh they're they're analogous to fox great states or you know we we sometimes call them squeeze states depending on how much uh how how much entanglement we've generated and the basic uh method of doing this if you want a nuts and bolts pictures we put a bunch of atoms at low temperature in a cavity and we measure how many atoms are in the cavity and and when we do that if the atoms are initially in superposition states then after that measurement we kind of create an entangled wave function that correlates the internal states of the atoms uh across the ensemble now there are a lot of details there but the bottom line is we we create this you know marvelous macroscopic entangled state and then what we do is we feed it into an interferometer and that interferometer in this case is configured to be an atomic clock and so if you if you think about you know some of the the gate operations in an ion trap quantum computer you know a lot of times what you're doing is is basically building a kind of complicated atomic clock you know you do pi over two rotations uh you do free procession and you know pi over two rotations you read out and what we read out instead of reading out single qubits is we read out uh the number of atoms in a in a given state across the ensemble so something simpler to do i don't have to detect individual particles i can detect ensembles uh but i can certainly tell in terms of the performance of the clock whether i benefit from the entanglement and so this is a clock engineers plot that shows the performance of the clock as a function of the amount of time you look at its output and basically uh you like to see these these uh these lines be lower on the on the on the chart uh to give you that shows it's a more accurate clock more precise clock if that's happening and uh you can see the blue is what happens what you build without entanglement the black is what happens when you entangle and you get us a substantial improvement in clock performance and so what we're working on doing now is translating this stuff from laboratory into uh you know atomic clocks that you you might find in the data center and uh there's lots of reasons to get a better atomic clock uh and again we can we can talk about why you you might want that uh the the de facto existing technology is this thing from uh i don't know if it's called microsemi now everybody keeps buying the companies that sells clocks but that's what it looks like and what we want to do is is bring entanglement to that beast and to improve its performance and so that's happening uh out at ao sense and a lot of labs metrology labs across the world are trying to figure out you know what the right recipe is to exploit this large scale entanglement for improved time standards and other sensors uh let's see one one interesting thing i said there's entanglement there and you say well okay yep what do you mean by that and and you know i can talk about the entanglement against a lot of different metrics uh you know like the noise the sensor would have if it wasn't entangled or uh the noise the sensor has when it is entangled one of the interesting metrics is to think about the bell correlations that exist in the uh in the ensemble this is kind of like measure of two gate fidelity in in in in in quantum gate ops and uh so back in 2017 uh uh a part of my team uh you know went and looked at a a measure of the kind of the existence of bell correlations uh in these ensembles and it's it's sort of it's it's stunning how how quantum the states we make with this this cavity assisted method are and in fact we observed a 56 sigma violation of it it's not exactly the bell inequality we it's the formal words are the bell witness criteria that's a formulation of bell inequality for macroscopic ensembles this just shows and that's what this chart shows is that's like this is this isn't your average classical ensemble of atoms this thing is is highly correlated and when we that's what we need to feed into this this measurement apparatus to in you know actually improve the precision of a cloth all right let me give you one last example and then i'll then i'll turn it over to you guys uh what i just described some some a couple measurement strategies that um were used to kind of you know look at things that were kind of macroscopic or you know uh can i use similar similar methods uh at the nanoscale you know building interferometers and so forth to improve microscopy and if you if you in microscopy uh you know and now we're getting you know kind of far afield from i understand from you know quantum hardware but stay with me here um in microscopy uh one of the things that you you'd really like to do is it's this is holy grail category i mean in my view and in the view of you know life scientists this is this is as much of a holy grail as you know building a well maybe not a fault tolerant quantum computer but building a good piece of quantum hardware that computes i i'd like to be able to look at small stuff without destroying it and right now i can't do that when i you i mean you look at the cover in nature and you see all these fancy electron microscope pictures you know most notably recently of the coronavirus but you know also also of other small proteins uh like turns out that when you look at those with electron beams then that's that's the only probe we have really to see at the nanoscale though those it's impossible to right now image that stuff without absolutely destroying it the radiation damage from the electron beam is is just monumental and so the question is can we can we do some quantum measurement uh in such a way that doesn't destroy what we're trying to look at and you know what like could we image something at the nanoscale that's alive yeah we have to just just be a little bit you know philosophical about what we call living you know but uh you know maybe i shouldn't put it in such metaphysical terms yeah like could we see a protein fold without without destroying it that kind of thing and that capability is enabling for example for drug discovery okay so uh what i just described with in that previous entanglement example was uh a quantum non-demolition measurement for atoms and i applied that non-demolition strategy to an ensemble of atoms to create the entangled state that i use to make the clock question being is there is our family quantum non-demolition measurements that i can also use for imaging small proteins and you know speaking like uh you know in terms of uh you know quantum computation here what you know is is it possible to entangle the state of my probe the electron with the state of entangle with with the state of um the thing i'm trying to image and then after i've done that entanglement then i can use all my quantum engineering tricks to collapse wave functions and so forth to learn in controlled ways about the thing i'm trying to study and you know on the on the right is a is a is one of these biophysicists sorry structural biology views of a protein um it's it's a virus particle uh and so we're building it turns out you can do that and uh what the the key ingredient is um when i use that that cavity to um non-destructively detect atoms what i ended up doing was taking the wave function of a photon and making it pass back and forth through the ensemble of atoms a lot of times and it turns out that that sort of strategy dramatically leverages the uh inelastic to elastic scattering ratios in the favor of elastic which is if you want to do coherent operations and you're a hardware person that's what you're looking to do and so we want to do the same thing with electrons we want to make an electron go back and forth through the image target in this case is the the marv vp35 uh protein which is something from my perspective as a physicist is something i just download from the protein databank and i have no idea what this thing does but i you know people tell me it's it's interesting to to to image that it's it's about five nanometers across and that in that image and i think it's 15 or 20 nanometers long and uh if you put this thing in a conventional transmission electron microscope you just melt it uh and you know kind of if you try and simulate what what you could do if you didn't melt it you would you just wouldn't have enough electrons going through that thing to see a darn thing and that's what i kind of show in that on the right hand side the picture there if you build one of these multi-pass quantum non-demolition microscopes and you make the electron go back and forth like dozens of times it turns out that you can actually build up enough phase shift as the the electron wavefront goes through that that protein in order when you when you make the measurement collapse wave function and so forth uh you can actually see the protein and in a quantum non-destructive way which is is sort of amazing uh is the only way this can happen is via quantum mechanics and so right now my group we're trying to build one of these microscopes i can tell you in a year or not whether it's going to work one more thing to tell you and then i'll i'll finish so uh maybe in your intro quantum of course you were introduced to the idea of a counterfactual measurement and i know there's you know there's some qc schemes that uh invoke counterfactual and it's it's it's it's one of these magic and mysterious things in quantum mechanics in my opinion so look at this diagram on the left it's a moxen under interferometer quantum particle goes in both paths at once hits a beam splitter at the exit and depending on the relative phase between the two beams the detector d1 clicks or the detector d2 clips now because it's an interferometer i can align it so that the detector d2 clicks every time just by adjusting the relative path length so i do that and so it's kind of an uninteresting thing i put particle in the input part and d2 clicks now do the following thing put an object in one of the arms of that interferometer uh maybe it's a piece of protein and i'm using an electron to do it and ask what you expect to see happen at the output of that interferometer well now if you think semi-classically i i blocked one of the arms so i no longer have that interference that made the electron come out and hit detector d2 and have a click every time if i block one of those arms what must happen is part of the time d2 clicks and some of the other times i send an electron through d1 clicks now if detector d1 clicks i have to have be able to infer that there's something in the interferometer because if there was nothing you know if there if there's nothing interferometer d2 has to click every time i put something in there the only way i can get a click in d1 is something something's blocking one of the arms but the detector clicked so the electron was not absorbed by the object that was blocking the arms it the electrons sailed right through that apparatus did not quote interact with the uh the object and and instead it interacted with the the clicking detector d1 so when d1 clicked i learned non-destructively about the presence of an object in in the arm i i observe that object if you will non-destructively so that's a that's a category of quantum non-demolition measurement and what i just described to you with that that multi-pass microscope is one where we want to change this is this is kind of a bad way of doing non-destructive measurement because some of the times d2 clicks uh even when the object's in the arm and and you can't make any solid inferences but by by doing that multi-pass method you can get to a situation where you're highly efficient uh in in this counter factual inference that when the your detector clicks you know that there had to be something uh in the in the um in the in the case of the the microscope the image plane last slide here so we're trying to do that with electrons but you know before we you can also light is also gives us marvelous quantum particles called photons so can we do something like this with photons and so we've we've been building microscopes that are exploiting that same counter factual principle uh and this is the same multipass principle we call it where uh you know when something in your detection plane clicks you you can make the uh the the strict inference you know kind of validated by quantum mechanics that there had to be something in the uh the image plane and so you know we we're we're just getting our proof of principles going on the left you see some of some some slides where each one of the bright spots on that is when we send a photon in and a single photon detector clicks and we can infer that there's a a small bead in the in the in in the uh in the sample plane and the right uh just for fun we put a red blood cell in there and and when we see a click on the right a single click we infer with with high probability that there had to be a red blood cell uh in the past where we want to go with this well still we're still trying to figure it out i think there actually may be quantum compute paradigms involved in this but uh it's certainly i think also interesting to to ask him can we do something more uh fantastic you know by in in entangling uh you know biological systems with with quantum particles and then and then uh manipulating and measuring so that's what i got and uh let me just cap you know so what i what i wanted to do is kind of expose you to if you haven't been exposed to you know some of what's going on on the on the metrology side of the quantum 2.0 revolution and uh a lot of us uh in terms of overall funding uh the metrology side is um attracting a lot of the fun a lot of funding i mean in the in europe it's about 40 of the quantum investment in the u.s maybe 30 percent and it's it's it's not as flashy you're not going to read about it as much on the like the cover of the the the new york times or you know the economist but uh you know the the reason why your is going there is because it's it's a little bit nearer term and uh you build a good enough sensor and you can do something that you you couldn't have done before which is sort of what we're trying to do with quantum computation and networking also so that's what i have to say thanks for your attention thanks so much mark um great um so i guess i'll open up the floor to any questions anyone has yeah we got one on the chat if you can't see that mark oh okay let me let me get up the chat here i'm not i haven't been teaching so i'm not as recently so i'm not as zoom-enabled as i should be so uh let's let's let's get uh i i can read it if you would like also yeah please do and i i um i'm having trouble getting the uh yeah okay this is what is that what i need to do oh this is from paul lipman mark do you think that this neutral atom laser cooling will be more likely passed to large-scale quantum computers than the super cooled qubit approach oh great question yeah and so uh out of all sense you know we're we're looking at this uh you know i would say um right now it's it's it's really hard to predict on the hardware side and uh they're quantera which is uh a company founded by some of my colleagues out in in the boston area maybe you guys have heard from him uh vladimir titch uh misha luken uh they so they're using neutral atom in lattice with rydberg gate interactions river media gates and they're they're actually doing some some that that system is really amenable to some quantum simulation tasks and uh you know they have they i mean they had some big splash results a couple years ago and uh and and now they're if you go to the quantera website um you know they're making 2d arrays of atoms and and doing kind of i think spin glass type quantum simulation work um where where is the bottleneck on neutral atom qc with rydberg well if there's uh uh like just like with iron trap uh you you have a spontaneous emission and inelastic interaction in your gate operation that uh limits the efficiency and fidelity of the gates um but it it's at a place where you know i think you can do intermediate scale i i don't think you'll get large scale but uh i mean arguably nobody knows how to do large scale fault tolerant right now but at the intermediate scale every hundreds of qubits uh the numbers sort of sort of pan out it's just it's it the the frustrating thing is and why you just don't want to jump in and say okay we're going to go all the way on this is it's it's not clear that you can you can you can extrapolate to a larger number of data operations um but it's certainly certainly very promising and it's it's just goes to show you on the hardware side that you know if you've talked to people four years ago and said well is anybody going to do a neutral atom lattice qc that's competitive with uh superconducting or uh ions um or solid you know solid state you wouldn't have you wouldn't have said that you know that's they had a chance and then then uh you have a disruptive idea and then the hardware starts uh you know coming along and so it'll be interesting to see where they go and there's another company up in berkeley uh adam computing that's doing this with atomic strontium uh quantera is an atomic rubidium strontium has some favorable level structures it maps onto precise atomic clocks and so a lot of the physics that uh you know i talked about with entanglement for improving quantum you know quantum clocks is is what they're exploiting up in atom computing so i don't know if i had to date i'd i put a bet on them mark uh actually that's very interesting this morning i was just finishing my teaching with explaining the elser vladimir uh bomb test for students and it's kind of realized that it's really cool if there could be practical upcoming uh tests or validation experimental uh to really show how the theory is matched with experiments do you know anyone trying to build some kind of equivalent process or something yeah for sure there are some uh let's see there there's there's some if you go into the literature on that there are some foundational experiments done in the early 90s with photons that you know kind of are considered like the experimental demonstration um for the high efficiency uh bomb testers the ones that you know are a little bit more than max enders uh i was involved in one of the early experiments in the in in the 90s with photons and just recently in fact this you know just this afternoon i received a paper for review that that talks about uh doing this kind of thing with uh electrons so you know it's it's those demonstrations are out there in terms of having real impact uh i don't think there's yet been a device on them on the measurement side that takes us beyond just like this is a quantum curiosity but i'm i'm optimistic that that that's going to happen primarily because of some of these biological applications um where uh you know in in in transition electron microscopy uh you have when you look at and build up these maps of small molecules uh you have hundreds of thousands of identical copies of the system you're looking at and so if you have a mechanism to know you know when when you're probing a system that you know the time you probed it it was a non-destructive probe then for the for those those on those subs those systems that you know you learn that about you you'll actually have a a damage-free inference a lot of the other examples will be just the bomb will explode and but as long as you can detect that you can hear the bomb going off you just choose not to look at that you post select uh so i think i i think we're going to see um micros uh evolving uh in this direction yeah it's fun it's fun to use that though isn't it the thing is that in quantum physics that's one of the achilles is here that once you look at something you change its state and if you want to maintain that you're continuing to operate on unchaste unchanged state then you have to actually focus on this kind of non-destructive interactions right thank you yeah thanks thank you hi mark you just ask a question can you detect neutrinos uh could we detect neutrinos uh there there are so speaking broadly and and maybe not as much in quantum measurement um you know language there are proposals for neutrino detections uh with with ensembles of atoms and it's up at triumph uh there's a group that's uh you know kind of focused on that um i don't think the numbers are really competitive so uh you know there's there's no large scale investment in that but uh yeah it's certainly been discussed and this is sort of an indirect path and i i'm sort of forgetting some of the details but i know there's a group at triumph in in in uh i think you've uh up in up in vancouver that's looking at that the reason i ask is um earth's interior with neutrinos oh yeah yeah yeah i'm interested in a really cool stylography it probably is better to use a conventional detector to start but if a really great uh neutrino detector you can let your imagination run wild in terms of applications and i've finished some environments on the like the national security side where where maybe people have let their imaginations run a little bit too wild yeah but math is snapping me i'm not that guy okay the mapping the earth mapping is is really a it's cool application yeah um that's the reason i ask thanks yeah you're welcome hi mark um back on the on the photonics side um i was wondering like psi quantum claims that in the next four years they're going to have a million cubits so on with photonic uh quantum computing so do you think that statement is believable like to what extent oh it's for sure believable they they they have the um the capability of doing the large-scale uh photonic integration and that that technology is there in the foundries what is but as as we've learned it's it's it's not just your qubit count and so uh you you can put down a million qubits and that's that's their sell is that they're they they have a way of getting to the uh you know the fault tolerant error corrected thresholds and cubic count that you need uh but a lot of people are wondering whether they can get the fidelity to where it needs to be and and uh in the and also the detection error rates and and so forth i mean you need to have that the photon comes in that detector needs to go quick and if it doesn't if it misses one then that's just going to make your uh your compute cycle more inefficient and so even though you had all those qubits you know you can't use it so it and and and you know stabilizing the interferometric pathways between uh you know million qubits uh you know i don't think it's i think even the cy quantum guys would have to say that you know they're just going to get into uh a regime that you know they're going to have to go into learning mode because these multi these multi-photon entangled states uh you know have just we can predict that they're super fragile and so you know it's it's a question of after you lay down all those gates whether you can uh you can actually use them to do something and what i find really exciting about sci quantum technology is there's you know a lot of one message that i've had in my involvement of pushing quantum technologies is you when you do that you're you're perfecting classical technologies and so uh there may be other uh sensing or compute paradigms based on classical fields that might also advance and so uh you know i i think it's gonna be really cool to see where that the psych quantum stuff goes uh in in the coming five years and it's it's a it's a very impressive effort i you know i think it's uh um i should have mentioned them in in when i was just rattling off a list of compute options because i think they're one of the dark horses out there and some they don't consider themselves a dark course they consider themselves a front runner but you know right now everybody you know you said what's the most likely i think most people would say oh it's going to be the superconducting qubits and uh you know i think when you take a sober analysis of where superconducting is and where it can go and where optical is and you know photonic gapes right you you start to be enthusiastic about photons great thank you all right guys well thank you very much again mark for the great talk and thank you stanford quantum for uh taking charge uh i guess you know we've had a very good start of the year uh a very good turnout thanks everybody who stayed with us almost two hours so for the next uh talk we we're pretty open right um we are doing kind of once a month so between you know me merc josh and maybe some other folks who um who want to to organize will get together uh will come up with a plan and we mail everybody um because everybody had to register so we have the email so we'll email you guys the the update and we're looking forward to really making it a fun destination so if you want to um give a talk please reach out to any of us i think uh you know we'll basically make it fun thank you guys yeah thank you everyone thanks everyone yeah hey thank you enjoy thank you thanks again mark and brad that was excellent you know thanks for starting this year on a great note so looking to learn more from you guys okay thank you bye bye