Quantum Conversations III: Hardware Overview and Investment Landscape

Quantum Conversations III: Hardware Overview and Investment Landscape

Recording: Quantum Conversations III: Hardware Overview and Investment Landscape

been to the lab since march uh but luckily i've been um able to and i'll show you this today i've been able to use uh the kinds of uh open source offerings we have kiss kit because kid pulse to be able to do experiments and and um all kinds of things you know over the internet instead of in person so it's great very good looking forward yes and and you will obviously get a chance to reintroduce yourself once you begin your talk for folks who will uh show up by the time uh so let's uh go with andre next all right i uh used to live in new york just like nick for 15 years but moved down to miami beach like many new yorkers about a year ago so i'm looking at the beach right here and i promise to only take three hours of my time and then i'll be going to the beach with my picture of margarita um i'm originally from germany i've lived in the us for 20 years my background is a split between a dozen years of management consultancy and then a dozen years as an entrepreneur where i founded five startups um then became a mentor investor and so forth over the last three and a half four years exclusively in quantum great thank you and again we're looking forward to talk i appreciate you joining us early uh you know i hope it's it's of interest but obviously you know we're online so uh everybody is is present from around the world uh now uh michael seder yes hi michael tell us a little bit about yourself and what are you working on well my background is electrical engineering i'm interested in all aspects of quantum computing and quantum physics uh and i'm located in um uh on the upper east side of manhattan new york city nice you know guys if if uh if you like please uh share with us what you want to uh to get out of these meetings and maybe what you want to do at these meetings if you want to present or if you want to find collaborators all right if you're looking for reviewers if you're looking for ideas uh so feel free to to add that because it's essentially the community meeting we you know we can best use it basically as as best we can and it's really up to us to have fun and and and basically make the most of it right so so feel free to add something uh so maybe michael you'll show like what are you looking for in this meetings what you know what's most interesting to you um you know well i i spent most of my career uh in the um in the computer industry i'm so i'm i'm just interested in learning uh new aspects of computing and i see quantum computing as the future of computing cool yeah i think a lot of us are here for the same reason so sounds good uh thank you michael uh taha jaffer hi folks can you see me yes you're well okay and we're here too yeah thank you very much um so i i'm in toronto canada i work for a scotiabank and i run one of the large ai ml labs that we have here and part of our mandate is on the quantum computing side in terms of bringing uh value uh to the bank via quantum computing in in the near term and so what i'm looking for from this meeting is is not just ideas and collaborators but i'm also looking for to make new friends as well as just to learn more about what's on the cutting edge and what's happening we have quite a few partners in and around the toronto area that are in the quantum computing space so you know we're trying to find ways to make this a reality instead of being something that's five to ten years down the road uh and you know so at this at the bank we're kind of the pointy end of the stick my background is um is in the hedge fund industry i'm not actually a data scientist i'm a quantitative portfolio manager so a lot of the things that the quantum computing site can give me i think is an advantage but on top of all of that i have a background in electrical computer engineering as well as physics and finance and math so this is all really right up my alley in terms of the kinds of interests i have not only about what's available today but what's there in the future and so joan was very kind enough to connect me into into this today so this is my first meeting and i'm just looking to learn more thank you doctor that's really great to hear from a fellow organizer now you know it may seem easy from the outside it's a little work from the inside physical virtual so and toronto uh you know we'll be here i have a lot of links with toronto obviously you know uh waterloo folks you know work all around the tech industry and yeah i think there is a strong community so my former cto actually just moved back to toronto where he was wrong right since bayer is kind of is all virtual now so great great connections uh thank you let's see uh so alex you're next alex kissinger hi i'm alex kissinger i'm a associate professor at oxford university in the department of computer science um i recently joined the quantum group there in computer science which has for a while been a more theoretical kind of quantum foundations group but something that i'm getting started there is a new quantum software activity there so so um starting from next month uh i'll be leading a team there of about six six seven people mostly phd students and some postdocs um thinking about things like quantum compilation quantum circuit optimization um and uh really doing things which are kind of geared toward uh the hardware that's becoming available today so so something that's quite interesting for me to be in a meeting like this is to meet some people also outside of academia and industry who are working on these kinds of things to find some new collaborations and and also uh some new ideas about joint projects maybe that the students or the teens could do um i also um gave a talk in the first of these meetings so so that's how i found out about this um and yeah i'm glad to be here so thank you alex yes and uh if you folks uh i want to review the talks everything goes to our youtube channel which is uh found at the easyurlfunction.tv so this is the youtube channel where we publish all the meetup uh conference talks we produce in the bay area and so there is actually more than a thousand talks i think about 300 sponsored by ibm from ibm i conferences uh and but a lot of it is basically community content from big data software engineering uh data pipelines so you know i think if you are into computing in different aspects i think you will find kind of a lot of cutting-edge content from all the bay area kind of usual suspects big and small right so and so this is all going there obviously it's you know creative commons free with attribution embedded you can share it right it's really educational resource so uh so yeah and alex was uh yeah i think the pioneer of this uh and uh we met him through through ibm so i'm i'm really happy you know how this this worked out and in terms of uh disrupting the academic routine i think we're all for it because so what you know what we find in in the bay area is that kind of if we disrupt the the kind of uh traditional resource of something uh fun happens not necessarily you know good for you know but but something interesting right and so i think what i found when i yeah so i went to a bunch of um quantum talks at berkeley when the world still was open and right and they were hiring i think for and this i think they're covering for two positions right and that that's between uh electrical engineering and physics and computer science right like is a lot of schools right they really i think uh they want to step up and they want to feel the faculty so you know i went to the faculty talks right where like the young scientists come in and they want to basically put a stake in the ground and so what's interesting to me i think that we're kind of on the verge of change because a lot of this young scientists are coming from this effectively in the environment you know all the like lots of collaborations are online a lot of startups are kind of uh like lurking around or like the idea of industry has changed right i i when i went to you know grad school you know like uh more than a decade ago i think kind of the real world was the anathema right like like you know real world will contaminate your science but i think in in computer science especially uh it is uh now uh uh almost par for the quarters right and so i think it's interesting i think quantum because it comes from physics i think there is a little bit more conservatism uh in academia but obviously it's changing as it's productionizing so that's why you know the idea of this conversations we want to really diversify them we we want to bring in folks not necessarily from from purely academic background we want to bring folks from industrial r d and and folks like andre who can tell us right where the money is going because in silicon valley we find that you know though money is sometimes uh kind of contaminant on the in other cases catalyzer right because a lot of smart people are thinking where the field is going they have this vast networks and they're talking to each other right as a jungle telegraph and they're kind of um generating a lot of knowledge filter and seeing what works what doesn't apply in this kind of pragmatic mindset and so in that sense money is a very interesting barometer right so so we really see um if i can get 60 seconds i would actually love to ask um our audience a quick question um how many quantum startups do you guys think there is currently in the world if you look if you think about it and across hardware software quantum sensing uh encryption uh shout out a couple numbers if you don't mind how many stars put it in the chat you guys can put it in the chat and i can read them per perfect and uh same question how much money do you all think has been invested this year so far globally private capital angels venture capital private equity to the closest 10 million um i'd i'd love to hear your thoughts that way maybe i can prepare specifically to that a little bit so meanwhile i'm going to be reading numbers andre so we have 200 320 and we have two numbers of a thousand so now we have a range from you know a few hundreds to to a few thousand to a thousand uh in terms of money uh yeah another hundred startups somebody give very specific number 314 companies maybe it's it's a pie i guess times 100 around it uh then we have 100 million and two billion in terms of capital estimates interesting thank you so let's see and yeah a 1 000 startups gets really a lot of votes uh so tell us are we vastly wrong at the like in the ballpark uh you're you're you're in the ballpark um [Music] all the guesses i think we got the two tail ends uh-huh interesting i guess you we need to come to your talk to learn the exact numbers absolutely sounds good sounds good uh cool so sorry i took a little bit of a detour with uh alex because i think he hit a lot of notes which which i kind of wanted to introduce as a theme ah and and we have our audience uh increasing uh so let's spend a few more minutes uh we have augusta kibowitz gilberts oh hi yeah i'm currently located at buenos aires argentina yeah i joined the last meeting i was working at jpmc and now i left the company but i really enjoy the talk so i i will plan to continue coming here i have a background in physics i did i did a master here in buenos aires and now pursue a more financial industry program a career but uh i i always loved the quantum and now that it's emerging and more accessible to everybody and the part of quantum and computing it's like merging all my my interests so i'm trying to learn a little bit and that's almost everything for me great thank you augusto this is this is awesome uh i love buenos aires and uh you know it's great city and convenient time zone right for the u.s as well uh the uh funny thing of gpmc i don't know if you're in touch with your colleagues uh it was interesting because you know we had a huge number of folks joining from jpmc at the uh previous uh uh meeting and ryden constantine is an amazing leader we know him for a long time but i didn't expect that you know he will bring such a vast group of colleagues so which was really great and kind of indicative the industries is interested however when the all registers don't zoom and we send them an email about this one they all got filtered by someone tech because they you feel the whole external email so this is you know another interesting feature right whole industry sometimes like all these kind of because they i don't know who knows i emailed constantine privately to tell him to share right ask him to invite others but this is an example of challenges we see because as a software engineer i knew that banks never use for instance java jars like they pick them apart and they look at them like you cannot just download stuff into a bank right like security will like maroon you for a week to to inspect anything and so this is an example how you know we see different um different groups right like moving a different speed so but this is great like you know you're out so we can you can hear like you could we can use you as a conduit maybe to like to relay secret messages back to gpmc uh welcome uh uh daniel mills uh is our next member daniel oh hello good to meet you all um yeah so i'm currently in london but i i work for cambridge quantum computing so um yeah we do a lot of things so we make compilers and we have some research in quantum chemistry and security so i work for the compiler team but i think these days a lot about error mitigation and things like this so approaches to making improvements in noise levels through other weight other means besides direct error correction um yeah and i'm just excited to meet with everybody quite diverse set of specialties here so i'm keen to discuss with you all what you're up to great thank you daniel uh and we got your email uh right we sent a request for presentations and danielle sent a great one and i should tell now it's very good that you shared you know uh your focus in in in this context because we we're planning for september uh we already have a submission on our correction i think it's a very important topic so maybe you know if uh if you want to preserve co-present right like we usually do two talks so that may be a focusing on that specific field because i think it's it's super crucial right for the whole for the whole uh area so i'll follow up for sure and obviously you know the proposals was great uh we're basically you know very accommodating uh and so we just want to hear from as many people as possible from different geographies uh and different uh you know like industry academia startups and and uh investments right all kind of fields related so please you know uh send us your ideas uh we it's just a question of time right because we're meeting monthly and we're trying to do two talks uh but you know we may do like special meetings uh in between so just you know uh if if you want to speak we probably will find find a way to do it and obviously we'll have lighting talks after this if folks still kind of paying attention uh thank you daniel uh jay mcgill uh everybody i am out of wisconsin and i've been catching as many of these type of talks as i can um so excited here from andre again and i hear from nick as well nick actually showed me the first quantum computer i ever saw when i was in ces he was kind of answering questions for ibm uh earlier in the year in january so yeah just uh really good group uh lexi so just looking forward to the talks today thank you jimmy yeah you know i i started recognizing the regulars so you know it's good uh it's good to see again uh and again like the order is weird if i read somebody twice please you know don't take an offense uh amir ibrahimi hey there i'm here yeah um yeah it's good to see some familiar faces here uh alex uh i certainly enjoy pi zx so thank you for your contributions to that and yeah i work for unity technologies in our machine learning and deep learning group on the inference side and just my main goal is to bootstrap until i'm doing active research in the field very cool very cool and unity obviously is you know a good partner so um awesome awesome to see this overlap of industries uh now john stanton uh hi alex welcome um thanks for having me again i've been on a couple of these sessions uh i'm from canada so we're here you know with the moose and the bears and not in silicon valley so in your honor i'm drinking uh coffee from san francisco cup today so on it anyway uh i do work with ibm here in canada um almost 45 years in a number of different areas but uh lately in the financial services industry and specifically do work with scotiabank so with some of his team and what we're really interested in was some of the information shared by jpmc and some of the others about areas where they're working so not just noise reduction but also derivative analysis portfolio optimization uh some of the other areas where people are starting to see some positive uh work as they move forward on this compared to standard state so just interested in in general and what others are doing so thanks again for arranging jpmc last time they were very good so looking really interesting chat today with the different speakers thank you john yeah i appreciate it i raised you a glass of verve which is santa cruz coffee you know which is uh also a great local coffee highly recommended thank you thank you it's a good good way to start today i know you are guys a little bit ahead of us and yeah that's i mean to me it's super interesting uh that so many financial folks are in this right obviously the the indus like technology's years away probably from from direct impact but uh i think banks are the most pragmatic and frugal companies and so if they're doing it i guess it certainly is of value right so to to so this is interesting to us again right because you guys are seeing something and you're obviously placing these bets and resources for a reason so it's really good to to have you guys here and and have you share it right to see how this is going so welcome uh now uh let's see terrell france uh i'm from i'm a professor at harrisburg university in central pennsylvania in the u.s and i get paid by the cia actually to to to follow jamie and amir and occasionally michael on weekends um in fact i thought jamie was in another presentation just 20 minutes ago but i guess he's here as well i think there's several jamies but uh you know i'm here i i'd seen nick talk or as well as andre several times i always walk away with something new new piece of information so i hope that continues uh i'm also here to this is my first conversations uh event i've seen you on the radar uh two or three times before and so i thought i'd get a feel for that and that's probably what i'm doing here uh anyway i'm new to quantum i'm new to quantum only i consider myself a freshman if you will uh i just been digging into the topic maybe uh 18 months or so i think we're all freshmen i don't i know only a few seniors probably i don't know if like anybody has a real phd in this probably some some folks do but like the field is young right so so uh so welcome and thank you uh terrell was sharing uh uh darrell has a resource actually right maybe you can put it in the chat for quantum events which is very thorough and actually covers a lot of things uh which i didn't hear about so i think it's a discovery and he obviously discovered this so uh and i guess you know the great thing about uh scientists they have the machinery or uh kind of uh filtering and absorbing new knowledge right and so i think it's great like a lot of folks in academia can can go into this new field and i guess this is how computer science started when computers appeared we know computer scientists or mathematicians physicists philosophers philosophers and linguists right in a way i think we see the second wave of this where people come together from different disciplines and and scientific apparatus and method and community is very important so in a way where you know we're building a collaboration framework for the same kind of new age so so welcome listen so this is great um uh what's uh vasillini chive for quantum computing and wanted to see what's happening in the field and no no people from from this year great welcome uh and veselyn gyorgyev uh he's our other uh i think uh regular uh member also from academia wesleyan hope you your connection improved i think you're a mute uh yes thank you so i'm also drinking from my cup that's actually associated with my previous uh position that was in china as a visiting assistant professor there uh but then it was 2017 where i realized that quantum is coming from h and decided to come back to california so i'm former uh kind of a academic career path uh entrepreneur that is switching to understand entrepreneurship last month i went through the cdl which is to run to quantum boot camp and i really enjoy it i highly recommend if you have a chance to attend some of the events that was fun uh from there actually i came up with a small team that we went last week through icorp berkeley and now we are waiting for uh their decisions where they were interested in what we are trying to do to keep going so i'm trying to build the startup for around quantum computing and its applications thank you hey wrestling where where in china were you uh a prof sichuan university yeah i was at beijing uh and how i was in i too left china i think i was about two years ago now so yeah well at the same time probably yeah yeah i was in shenzhen i love china i mean they have very interesting culture and interesting people you know you know the funny thing is i lived in shenzhen which is like the electronics capital the planet right and and i was in the business school there uh at peaking university and i moved back to the united states and pivoted over to quantum to cyber security iot stuff and then quantum but it was ironic you know i literally lived you know a few kilometers away from these factories that make all these iot devices and i never bought one until i moved to the united states now you know then then i just binged on it so everything was coming from my neighborhood where you know when you live there it's like you just pass it by it's like the statue of liberty in new york city you just don't pay attention until you leave then all of a sudden uh you know i don't know how much i i probably paid you know 20 for something i could have bought for 50 cents in shenzhen that's true interesting uh all right so our final member uh to introduce is jiong zhang welcome to young he's a regular for all quantum conversations hi everybody um my name is julion john and thanks alex for organizing this event it's very interesting i have been picking on regular now it's kind of amazing it's early in the morning i i'm a quantum chemist and communication chemist and i work at stanford research computing center and my interest is applying um quantum algorithms to computational chemistry applications so that's what i have been trying to do so far thank you great uh welcome jiang again uh all right so i think uh we actually went through all the introductions uh and uh alexi it worked real yes uh there's there's there's one thing i forgot to mention if i had have maybe just one minute to say one more thing um so my soon-to-be phd student uh leah yeah um has started putting together a lot of open source quantum computing educational resources online which is something some people here might be interested in and also runs a discord server with i think by now some 50 or 60 people kind of working through these things and doing reading groups and this sort of stuff maybe if i if i send you the link you could sort of bounce it back to everybody else over the chat because i don't i don't think i can broadcast something to everybody on the yeah i think uh there were some instances of zoom bombing and and we had together to restrict uh all this this works so yeah please please send me the links what i'm gonna do i'm actually gonna publish them on their uh our homepage quantum dot sv which was originally meant to be silicon valley but obviously now it's you know the world right uh oh yeah please uh that that's great the question is um uh uh are these resources used in that you know in your work because if you know uh one way you know to to have more fun would be maybe to walk people through them or have a discussion of having some kind of hands-on tutorial so so obviously when folks look at open source if they use it they play with that it's still fresh in their minds you know how it works like sometimes even installing open source takes time right and so uh i would propose like i i'd love to have a talk where somebody installs a you know open source package shares what's happening right so it doesn't have to be oh a slide presentation actually in in kind of software engineering startup world the best presentation is live coding people just called and like they're in the terminal they're in the notebook uh like you did right so like this is this is you know this is uh very engaging and i think it's for a lot of people so i would say you know uh please you know consider kind of you know giving like a live coding talk of some sort reviewing this or you know something highlights right so we definitely welcome it yeah maybe maybe i'll uh i'll see if leo wants to wants to give a lightning talk at some point in one of these to explain kind of what's out there and and and how how this community she's setting up is working and and and so on so um well anyway the address is full stack quantum computation dot tech i i sent that to alexi um that'll that'll appear on the website and maybe it'll it'll appear in the chat if he if he repays that or if one of the hosts does if i figure out how to broadcast it because i think i locked myself into a corner now and i can only chat with people one-on-one so it's it's really and the funniest thing about when things are on the way like it's too late the boat has sailed right the train left the station so uh so yeah i'll definitely share figure out i see the link so i'll share it uh by the way let's say if you go in the shield down there you may change this capability if you look at the shields options uh-huh yeah okay i mean i'll be playing with this as i go yes i see the shield that's right that's right that's good that's right maybe i'll i'll see if i'll change it uh and i'll i'll i'll try to replace that as the talks progress uh and by the way guys this is i think thank you alex uh obviously you know folks have announcements right uh feel free to to make a brief for now because we have lightning talks for kind of sharing more uh after the main program but uh if you have any announcements feel free to to like alex to mention any any of those everybody and uh we'll do again you know we'll have a chance if you if you can think of an announcement we'll we'll probably do a quick round in between the talks uh right so and obviously we'll have some time in the end for those who stay all right guys i think uh that's good so now i think we're ready to start our first talk uh it's by nick braun of ibm uh and and he will do an overview of quantum hardware which is obviously the the key topic none of this is possible without hardware and it's much easier to play with software you know and you need to obviously be in a special place you know and i think a few companies like ibm can actually do this so we're very excited to have uh nick braun from ibm with us to give us an audio of hardware for quantum computing and some topics in that space uh welcome nick all right well thanks so thanks a lot for that introduction lexi can everyone see the slides yes uh and feel free to ask me questions um during the talk like i'm i'm happy to have a conversation and stop on you know anywhere before we go on to other things and we get lost again um so as you're giving the background i suppose i give you uh my background so i uh i'm an experimental physicist i did not study quantum computing for my phd because it was still very rare in 2013 when i graduated for people to be doing that experimentally uh it's a bit more common now but i i developed a lot of the same skill sets in particular you see right here this is something called a dilution refrigerator this is what gets our qubits very cold i was working with superconductors in my phd so a lot of the things that we use for building quantum computers out of superconducting qubits i was already familiar with so i came to ibm in 2013 started as a postdoc you know converted to a quantum computing expert and then became a research staff member which is where i am today and uh so today what i'd like to talk to you about is uh enabling near-term noisy quantum uh hardware which is by what i mean is i'm not going to be talking about air correction at all this is kind of the state of the art for what we have in today ibm we have a lot of these systems deployed that you can access online other companies have some as well uh and in particular i want to tell you about a more experimental way of accessing them other than kind of the gate-based circuit model that you know and it's kind of the more experimental angle that uh i think improves a lot of um improves a lot of capabilities and and i i hope to show you why that's important today so a lot of people go back to uh this slide uh where the ibm and mit hosted the first physics of computation conference in 1981 and if you look over to the kind of middle middle right back you'll see richard feynman and most people say he invented quantum computing because he wanted a quantum a computer that obeyed the laws of quantum mechanics to solve his quantum mechanics problem uh that would make it computationally efficient for him but ibm was already working in this space right now in fact uh guy ralph landauer who's in the front in the middle left and uh charles bennett who who is actually taking the picture here uh they have been thinking about the physics of information for some time before then in particular uh there was a lot of concerns about say the thermodynamics of computation because there's so much energy cost is there a finite amount of energy that gets spent and it turns out for classical irreversible computing there is it's called the von neumann limit um sorry von neumann landauer limit so there's always a finite amount of energy cost due to the erasure of that information but what a lot of these uh physicists did basically is they thought about well what are the physical properties of what i'm storing information in so a lot of times i'll make an analogy to classical mechanics in order to kind of realize what the difference is for quantum mechanics so if you think about a classical bit it always always going to maintain its uh its information in a physical system such as the on offer of light switch whether current is flowing or not through a transistor or whether uh or what what the polarity of tiny magnets inside of your hard disk drive are uh one of the one of the features of all these systems are they are very non-linear and they have a latching mechanism which means they'll flip to either the on or the off state and so their logical states are represented by this yellow dot and as a zero or this blue dot as a one okay so how's this different than the quantum mechanics well in um quantum systems physical systems such as the atoms or the uh or the spin of an electrons you can store the quantum you can store the information uh zero one in the quantum state of these kinds of things such as the ground or excited uh state of an atom the up or down uh spin of an electron or the the oscillations in little superconducting circuits that we use and if you uh so if you have uh information stored in this state your zero and one states essentially obey schrodinger's equation and because it's a linear equation you can take any superposition of those two uh solutions and provide another valid solution which is why we have something called superposition and it it it's represented by something we call the bloch sphere where we we have any point on the surface of the earth is as a as a valid quantum state where zero is the north pole and one is the south pole and then any point there is a complex superposition of zero and one um so what's interesting about this quantum state is these are all the valid solutions so these are the way you can encode them um it's different from analog quantum computing because we can also entangle the quantum states with each other which allows us to go to a much larger hilbert space for quantum quantum computation uh but what's interesting about quantum mechanics is even though we can put the qubits themselves in a definite state as represented by this block sphere whenever you perform a measurement you are going to collapse it into the uh the eigen basis of what you're measuring so for example if i'm looking at the z-axis i can only measure a zero or one and the point at which it is on the block sphere will give me the probability in which i measure that so for example any point on the equator uh after a measurement y will yield the the answer zero the answer one with 50 probabilities each so uh what i'm going to talk about now is going in specifically how do we encode this in superconducting qubits and how do we can control it from there so uh let's look at a five qubit superconducting uh quantum processor that was one of the first ones we put online uh four years ago and uh if we look in these square pockets we find the qubits themselves and the qubits are something uh that is pretty much just a basic superconducting circuit it consists of this 90 degree rotated equal sign which represents a capacitor electrically and it has this really small part between it called a josephine junction which is actually a non-linear inductor it's really really tiny it consists of a superconductor insulator superconductor sandwich it's the only non-linear uh nonlinear dissipationless inductive element known and and so essentially when we combine these two elements we have an lc oscillator albeit a non-linear lc oscillator but that's also important as well we also have superconducting microwave resonators that connect the qubits together uh and connect the cubes to the outside world now these are more like lc oscillators they're distributed they're not are not lumped like the like the cubit is but these are are used in conjunction with the qubits themselves to read out the states of the qubits uh to to interact the qubits with each other and they also are detuned so they function as a frequency filters at the qubit frequency so what's the big difference between having a linear lc circuit and a non-linear one well if i was uh if i was trying to store quantum information in a in a harmonic oscillator what you'd see is that you have equally equally spaced energy levels which means when i try to drive between zero and one i will accidentally drive between one and two two and three three and four so i cannot distinguish my zero in one states uh if i have a lc oscillator however i have an harmonic oscillator with a non-linear inductance then what happens is the energy difference between zero and one is different from that between one and two and two and three and so on so by just focusing um radiation at the transition frequency between zero and one i can pretend like i have only a two level system and i can do computation there so we can we also refer to zero and one as the computational basis now uh for superconducting circuits like these they operate in the microwave regime our superconducting circuits are typically around five gigahertz and if you multiply five gigahertz by boltzmann's constant divided by planck sorry multiplied by planck's constant divided by boltzmann's constant you'll see that that corresponds to an energy of about 240 millikelvin so that's substantially less than a single degree celsius above absolute zero and it's also a lot less than the superconducting critical temperatures of both aluminum and niobium which are the superconductors that we use those are about one kelvin and nine kelvin respectively so uh this is the reason why you have that fancy uh fridge on the uh on the first slide because we have to actually operate these these guys that uh up at a point of this fridge which is about 10 to 20 millikelvin which is about what you can reasonably get buying a commercial uh dilution refrigerator so i'm going to tie this back to what what we're actually doing mathematically so i'm going to go back to the block sphere for a second and i'm just going to tell you the computational states you might remember using direct notation which is the physicist's favorite way of telling you something is quantum just is a shorthand for these column vectors 0 means the column vector 1 0 1 means the column vector 0 1 they represent any points on the block sphere then i have these special matrices called poly matrices that were originally uh developed in order to describe the physics of spin and in in particular they provide the axes that you can measure across so you can measure along the x-axis the y-axis and the z-axis for superconducting qubits at least of our kinds we're always measuring across the z axis which means we can only measure these zero in one state and this tells us that the zero and one states are the eigen states of the system with eigenvalue one and negative one respectively but because they're also uh they're also projectors along those axes enough sorry not projectors but they also generate rotations around those axes uh and that's how we generate the unitary transformations that uh that that manipulate the states of our qubits for example so uh what you can do is any uh normal unit vector in the block sphere given by little n and you want to rotate an angle theta around that you can generate the unitary with this r and of theta here just by exponentiating that that uh that unit vector with a vector of the poly matrices for example and you get this somewhat simpler formula that's actually easier to use for calculation and so you can see what happens if i just want to generate a simple rotation well if i just want to go around the x-axis i throw the x-poly matrix in there and i can get out a generation uh the unitary that will move my qubit from the one state to the sorry from the zero state to the uh zero plus i1 state uh in particular for theta equals minus pi over two so this is essentially how all these qubits are how all these qubit control things work is is the unitary matrices are doing the operation on the uh qubits themselves so how do you actually do that in lab well like i said before we're using microwave pulses so we take a pulse at the carr we take a microwave pulse at the resonant frequency between the zero and one states of the qubit that's our carrier frequency and we generate usually a gaussian type something close to a gaussian which i'll show you later pulse that essentially brings the qubit between the zero and the one states and and all the superpositions thereof and so i can use these to do rotations around the x-axis rotations on the y-axis or in particular any rotation in the x-y plane of the block sphere the only difference being the phase of that uh carrier frequency is is what's defining the axis of the of the uh the axis of rotation in the xy plane um so that's great so what about z rotations well you can come combine x and y rotations to to generate zero rotations but if you realize you can actually do zero rotations for free because what happens is if you have a zero rotation in the circuit you can just compile it down and shift the subsequent um phases of all the other subsequent uh circuits and so what that what that means is it's something we call a frame change essentially it means that z rotations go away in software you can compile them down so they're free you're not actually doing a physical thing so there's no error associated with it and uh this kind of shows that you get an average single qubit error rate that's lower if you if you employ these frame changes versus actual pulses so then uh that's for single qubit control then what's special about quantum computing is is uh is entanglement so let's go to what we do with two qubits and i i heard there were some uh electrical engineers in the uh audience and so i i always like to go back to this uh i have a little bit of an electrical engineering background too i like to go back to this idea with classical classical computing is what makes a classical universal computer and that's the existence of being able to do any boolean operation and so you have two different uh two different gates that uh that will realize any classical system and that is the not and and the not or and essentially they you have these truth tables and if you combine those with ancilla bits then you can generate any boolean function that you want to and you can take electrical engineering classes that tell you how to do that the situation for quantum computing is very similar you need a certain number of single qubit rotations even though i just showed you how to do all of them um but you need a partic particular two qubit operation that generates enough entanglement and there's a there's actually a few different ones of these there's something called a control knot uh where the first qubit is the control qubit the second qubit is the target qubit and what happens is the uh the control meaning the control qubit will only flip the target qubit if the control qubit is in the one state so if i'm in the zero state zero zero goes to zero zero zero one goes to zero one but if my control cube it's in the one state then i flip the target qubit so one zero goes to one one one one goes to one zero and uh you can see how entanglement is generated uh by just looking at these truth tables because i can take a superposition of these input states say zero zero plus one zero and then if i do a c naught to it then i get the i get zero zero plus one one which is a special in highly entangled state called a bell state which we'll uh come back to later but that's not the only two cubic gate i could use i could also use something called a control phase and all a controlled phase does is it gives you a 180 degree phase or a negative one when both cubits are in the one state uh there's another popular one called the i swap which swaps the state of the two qubits and then if anything is actually done it gives you an i or the imaginary number the square root of negative one in front of it which corresponds to a 90 degree phase um so this is essentially the you know the the mathematics of why of of how this uh how you get the universal quantum gate set uh what happens in in the actual laboratory yes our question uh yeah i have a question sure um what happens if you apply a two-qubit gate between qubits which are in different classical frames oh assuming you like the by by which you mean rotating wave approximation um i'm yeah i mean so so when you said the the z phase comes for free yeah yeah then say i apply a z phase gate to qubit one um and then say i want to do a c naught between qubit one and qubit two um can can that be done just by bookkeeping or do i need to do something to sort of bring those into the same frame i see what you're saying um they are done by bookkeeping and i i think uh i think what i have later we'll we'll answer your question maybe we can address it at that point but that's yeah advance that's a that's a little advanced for this point but but good question um i have another question actually it's download me just now because this is really nice slide summarizing classical computing and uh quantum computing and i just realized that it seems to me there is no reason not to have a c not in a classical computational framework isn't that right uh you can have a c naught in the in a classical way the problem is is uh and this is actually you just made an insight that i i totally skipped over but it's very important notice that for the classical computers you have two inputs and one output whereas the quantum computer you have two inputs and two outputs and that's because quantum computing is a reversible theory or a reversible system of computing where classical computing in this way is not reversible there are ways you can you can configure it to make it reversible and that was kind of a lot of the ideas that uh charlie bennett and ralph landauer are kicking around in the 70s and 80s if you have a reversible computer then you don't have that von neumann uh landauer energy cost for example okay but there are yeah there are schemes in which you can realize this but it turns out not to be the same one that we're using uh and then i guess one of the other things is it's it's interesting because in quantum computing i didn't touch on this but you cannot copy arbitrary quantum information because you essentially destroy it by measuring it so you cannot uh essentially make a clone of it however you can you can't teleport it it's kind of they seem like they're opposite things but they're really or they seem like they'd be similar things but they're really not um yeah so continuing on so uh the choice of which one you use here for your two qubits is is kind of up to your qubit architecture uh even for superconducting qubits you can realize any one of these and then other other hardware is going to be more conducive to other kinds as well but for your different kinds of circuits there's different kinds of engineering choices that you make and so what i'm going to talk about is what we do at ibm we do something called the cross resonance gate and um essentially it's it's locally equivalent to a c naught you need some single qubit operations but essentially it's a zx operation and so let me tell you what that means so a zx rotation just means i'm going to generate a rotation that's uh based on the the tensor product of the z and x just like we we did before with the rotations in the block sphere now i'm going to take the tensor product and that's going to i'm going to form the rotation that's based on that so when i stick that zx in that formula i get a matrix where you can see that the rotation of the target qubit depends on the state of the controlled qubit which sounds a lot like what a c naught is and that's because it is and if you look in the matrix the upper left of this part which corresponds to the subspace in which the control cubit is in the zero state you'll see that this is a rotation of the target qubit around the x-axis and if you look at the lower right block you'll where you're in the subspace in which the control qubit is in the one state you'll see that this is a negative rotation around the x direction and around the x-axis in the opposite direction so this essentially just generates this rotation where uh the rotation of the target depends on the state of the qubit and i'm going to try and show you a picture to try and try and make this more sense so so whenever we think about things in physics we we look and see how things are done one with one state we look and see how things are done with another state and then we know we can take a superposition of any of those and that's valid as well so that's kind of how we think about it like in in this way and what i'd like to think of as just forget everything about the purple uh the purple vector right now so i'm looking at the control qubit in the zero state and i'm going to start with the target qubit in the zero state as well but what happens when i apply this cross resonance operation uh what what will happen is because the control qubit's in the uh in the zero state my target qubit's going to rotate to the right in this direction okay so that's cool well now let's forget about this purple cupid in the zero state and focus on the red blue qubit in the one state okay now because i'm in the one state i start with the target qubit also in this or sorry uh starting from the zero state we always tend to start from the zero state and then i apply the cross resonance gate then what happens in this case is my target qubit rotates to the left as represented by this blue vector right here so then in order to generate entanglement and this is something incredibly hard to draw i will prepare this yellow vector which which is the control qubit at an equal superposition of zero and one and when i'm in this yellow vector and i apply the cross resonance gate the target qubit starts on the zero and it rotates in a linear combination both to the left and to the right which is you know kind of impossible to visualize but this is the kind of the best i'm i can do but you can see that like this the superposition of both those states is is occurring at the same time okay cool so what's actually happening in the lab so one of the reasons that we choose the cross resonance operation the zx is related to c naught so that's that's very good but there's a lot of engineering reasons quantum engineering reasons that we're doing it the first one of these is that it's an all microwave entangling gate we do not need any flux pulses any kind of dc lines to to modify or tune any kind of element of this everything is fixed and that's that's the way we we do things at ibm because uh in order to get the highest coherence for your qubits we want to reduce the number of available channels for decoherence so if you keep everything fixed that's kind of one of the best ways to do it we think so we prefer this gate where you don't need any other tunable elements of course this presents a whole slew of engineering challenges that that you know experimentalists are working on the lab and everything i'd be happy to discuss later but basically i just want to give you an overview of what this is essentially we have one of these uh microwave transmission lines we call it a bus resonator because it couples uh quantum information between two qubits one on the left and one on the right the j is just the the coupling rate so they're slightly hybridizing the qubits so the left one becomes like the right one and the right one becomes like the left one so normally you would control your q1 with uh with pulses at the frequency of q1 in blue and normally you control q2 on its own with the pulses of the frequency q2 applied to q2 as well the frequency of q1 is applied to q1 so the reason we call this cross resonance is what we do is we drive the control qubit at the frequency of the target qubit and we apply it to the control qubit as well and so what that does is it generates rotations uh in the of q2 but at the dependent on the state of q1 and if you're a physicist like me you feel a lot more comfortable looking at energy level diagrams so let me just briefly explain what this is um so this is the these energies the energies on the y-axis here and uh i have the controls the first qubit and the and the target as a second qubit so if i didn't have any interactions between these i would have uh you know just uncoupled uh 1 0 0 1 and 1 1 energy states where these guys are going to be different from each other because i i didn't mention it when i when i said it but this joseph injunction when when you fabricate these things this is just kind of an aluminum oxide it's just kind of a you know amorphous glass that is between the superconductors so even if you prepare them all identically because it's an amorphous growth you're not going to get the identical uh critical current through these adjustment junctions so that's going to affect their frequency would be a natural spread uh so what happens is if they're not interacting they're they're on these dotted lines and as soon as you put this j uh microwave bust resonator between them then you've you've coupled them together and they're going to repel each other in energy and so you get this plus and minus this plus or minus j over delta so this kind of shows you the hybrid way where j is the strength of the coupling and delta is the detuning or the frequency difference in which uh they are at and for this reason uh you get this dependence on the state of the control qubit when you're trying to drive the target qubit depending on whether it's in the zero state you're driving around this one or the uh excited state you're driving around this interaction so that's the physics point of view and let me just say if there's any questions on that this is probably about as hard as it gets okay so let's see what it looks like in the lab and in fact i'm going to show you a little cartoon of what it looks like for measurement pulses so i've got a computer that's wired to a bunch of microwave instruments that goes into a dilution refrigerator and then back so these are going to be measurement pulses i'm sending in here this actually occurs at a different frequency usually around seven gigahertz instead of five gigahertz um and i can generate these pulses which just look like sine waves and send them into the fridge and so they go down these lines and thermalize really well and you know go through filters and stuff and they come to this uh did it freeze for everyone or just me yeah it's frozen whoa the error is moving up okay no is it going again now all right so uh so it goes down we've got to we'll thermalize it it's you know there's an art form and then these resonators will go in and they'll collapse the state of the cubit to the zero the one and they'll get a state dependent phase shift and part it on those pulses and so then those pulses return they're amplified a bunch because they're they're very small when they get to that point they get mixed down to a lower frequency that can be digitized and classified and stuff like that and i'll kind of show you how how this is done but that's the general measurement process and that's you know roughly what the lab looks like so uh i talked about these being noisy uh quantum systems so what do we mean by noise and what do we mean by how do and how do we judge the level of noise of these systems so we're going to tell you about little device level metrics and we'll get into a little bit more holistic one later which is it's something new i've got for you today um so as i told you they are kind of like analog control on these uh on these qubits so you can imagine when i'm rotating my cubit around the block sphere i can easily get uh over rotations and under rotations uh such as and here so if i do a single pi over two i'm off by some error epsilon and whenever i try and get back i just get two more epsilon each time i try and do it in fact we use this for calibration we call this an error amplification sequence because each one of our steps is going to amplify the error and then we can use that to more finely calibrate our systems but there's always going to be some amount of control error the other thing we'd like to do is i'll talk about the coherence in a second but because there's a finite lifetime of these qubits we want to try and do these operations as short as possible the problem is if you make them too short you get too much spectral content at those higher order energy levels so you can get leakage into the what we call the non-computational bases which are the ones that are not zero and one but say two for example you can also get measurement errors and in this error in this area i showed you what's the the iq plane we call these blob plots this is essentially the complex plane for measurements and usually we'll get a blob corresponding to one to zero a blob corresponding to one and a blob corresponding to i guess other things and then you can also see that when you have these blobs you have to figure out a way to classify them so here we've just kind of done it done it stupidly and put a line where everything on the left is zero and then put another line and split between one and two but this is always going to be uh a problem to classify whether you measure zero one because even if you had a perfect system you always have quantum noise of the measurement and you're always going to have noise of the amplifier so that's why these blobs get you know kind of big uh another thing we have is crosstalk and that's when we are operating on say two qubits and it affects the state of another qubit so and this is an example uh from us as well when we're trying to do cross resonance uh between two qubits but we have something called a spectator qubit that's sitting around but it's being manipulated by it and in fact what you're seeing here is a is a shift of the spectator cubits frequency and this frequency shift actually induces a z rotation error on that qubit so so that's also not good uh and then of course we have the decoherence methods we have we call it relaxation t1 so this is kind of more the classical one where you prepare the qubit in a state zero and it's going to relax back to the state one and this can happen several ways in fact you can have the opposite way if your qubits are hot you can have excitation from the zero state to the one state um but the main ways you know usually we're concerned with relaxation and this can be like spontaneous emission due to the electromagnetic environment so it kind of behaves just like a uh rc time con or sorry lc time constant in any electrical microwave circuit uh you can also couple to two level systems which are or any other channel two level systems are usually the material defects dielectric defects that exist on our in or on our device uh and then there's other superconductor phenomenon that can affect you things says quasi particles which are unpaired quick repairs cooper pairs flow without resistance but quasi particles do not uh you can also have vortices which can have their own resistances which are little tornadoes of charge that exist in some superconductors so there's a lot of you know very interesting physics that's happening these things that can cause relaxation and then there's another uh another thing called we call decoherence or t2 which is kind of an overall metric this is a combination of pure dephasing which is the moving along the longitude in in some uncertain way and this is typically due to other other effects than t1 but the decoherence t2 is caused by a combination of both the dephasing and the relaxation and this is typically due to fluctuations in energy levels that could be due to thermal broadening or magnetic noise if you have a tunable qubit uh whenever you have an whenever you have a frequency shift essentially you pick up a z air uh which which will give that you this dephasing uh whenever you intentionally measure if whether you measure intentionally or not you're going to collapse the qubit state so if you if you're not shielding from noise you could be unintentionally looking at your qubits while you're trying to operate on them which will cause decoherence and then ultimately it's going to be limited by this t1 and and this is kind of the i would say the overall metrics we have lots of different experiments that we go through to um to like kind of classify these uh and so i'm going to talk about one more holistic metric uh later but this is kind of it for the physical systems and so this might be a good point time for questions if anyone okay i'll just keep going on all right so what i want to talk today is uh kiskit pulse so how do we control how do we do all these things that i was just telling you about with uh with kiskit which is uh well originally it's stood for the quantum information science toolkit and so we shorten it to kids kit and kiskat pulse is part of it so just me just give a brief overview of kiskit so down here at the very bottom we have the quantum hardware which i was just describing to you the all the things we controlled how what we're doing in the lab and then we have kiskat there's a software that sits on top of it and so uh it's an open source software it's apache 2.0 which you use is python it can control non-ibm hardware it can control simulators it can control a lot of things so we try to make it as modular as extensible as possible and build tools on top of it so fundamentally terra is kind of the found the foundation for us to be able to run quantum circuits on our our back-ends or the hardware that we have in the lab and we have several of these that are open for everyone and several of them that are and then the rest of them um the premium devices are open for our clients um are all accessible via kizkit and then there's other systems that are available they're accessible via kiskit as well um we have ignis which are tools for for characterizing and mitigating air so this is kind of a you know a very popular thing because mitigating air is in order to get the best answer out of the quantum computer you're going to need to figure out how to you know do a lot of tricks and air mitigation is one of them i'll talk about uh we have air which is special purpose simulators try and figure out where my noise source is coming from i can throw them into you know some linblady and master equation and see see if i can predict what's actually going on in the lab or seeing what what kind of algorithms i can have that might be immune to certain kinds of error and then aqua is kind of the uh libraries for of quantum algorithms that like you might want to do like chemistry ai finance things that you can use and not have to be a quantum computing expert in uh so that you can start using you start using quantum computers to do uh whatever you want to do so an easier way to explore and these are all very you know we try to make this as easy to to install it's hosted on github you can do just pip install kiskit we just had to release uh 0.20 which i think has a lot of improvements we have a slack community we have an open source textbook we have um you know a lot of a youtube channel featuring we also have a youtube channel featuring a lot of talks um at the research level to the like coding with kizzkit level but here basically i want to talk about pulse which is part of terror which gives you this level of control so uh this is the second paper we put out about pulse i think it's just about to be published for real um and so this has been established since 2017 but but this essentially defines how do you want to how do you program quantum computers through the um how do you program like i was like if i was in the lab which i can't be because of covid but i can do it from home using uh keskid pulse and how do i how would i do that and how does it compare to what i could do in the lab and and the answer is there's a lot of things i can do with kiskit pulse that allows me to do things better and those can be things such as optimal control where you use knowledge of the hamiltonian and maybe machine learning or something like that to reduce the amount of leakage say outside the non-computational basis to uh in order to get faster gates this is one thing that's done error mitigation is something i'll also talk about and uh dynamical decoupling which is kind of a quantum way of eliminating um eliminating coherent errors in your system and these are all things that you can introduce with pulse and and and then kind of uh i had made the slide before and what i realized is that uh last week uh this is like very very fresh news we achieved quantum volume of 64 on our systems and so let me just take a step back and tell you that quantum volume is this holistic metric that we are using that essentially has to do with how large of an arbitrary circuit can we compute using our quantum computer so we just published this and announced this on thursday and uh this is essentially an improvement to quantum volume it was 32 before it's 64 now uh where we're using only control and compilation to sorry question okay maybe it's background noise um so this was made this was not with with any hardware improvements this was using the same hardware we had quantum volume 32 on but it was made by improving the kisket compiler and then other other methods i'll show you called a excited state promoted readout which means you can actually try and push the qubit from the one state to the two states so you can get better separation and classification between the zero and the one or what was the one uh we have shorter two-qubit gates which uh of course improves coh improves because of coherence and we're using dynamical decoupling and i'll kind of show you how this comes in at like a very low level um so basically this is the idea of near-term quantum devices is you have hardware solutions and control solutions to all kinds of constraints so you might have gate and measurement errors because you you always will until you have error correction so you can try and build more robust qubits or improve the microwave hygiene between your qubits or come up with a better frequency frequency scheme if you can which is difficult but you also have control solutions so you can try and calibrate your your measurements better because that's kind of a classical thing you can you can calibrate for you can use um say pulse shaping to uh to to get rid of crosstalk by doing say dynamical decoupling uh for other examples like hardware gate sets you can you can try and make more interactions like at a chip level like you can couple more things together but you can also do things a lot of things with pulse shaping um also the number of parallel operations you want to run a reduced crosstalk uh this this is something we achieved in software just using dynamical decoupling that's essentially done by pulse shaping so all of these things have a hardware solution that of course we need to continue working on but now i'm going to present today the control solutions um so basically what programming and pulse comes down into is you have four kinds of channels uh so you have a the drive channel and the control channels that actually control and act on the qubit the drive channels would be at the cubits resonant frequency the control channel would be uh off resonant drive which is you know often used for cross-resonance gates but it could be some other kind of two-qubit gate you could do then we have measurement channels that's at a different frequency and that kind of comes at the end when we want to measure the qubits and we have the acquired channel which you don't actually apply pulses to but that tells the digitizer when to start taking taking raw waveform data back and then we have certain instructions we can do we can play waveforms on the channel typically for the pulse channels we can delay channels so you can put time between the pulses which is actually very useful because as an experimental physicist your bread and butter is to come up with pull sequences and vary one parameter and see what happens and fit it to a theory so you can figure out what's going on in your system you can also shift or frame change and i'll get into this again but this this is what changes the frame between the qubits and if you keep everything you know together then you don't have to worry too much about it your acquire channel just tells you when to start acquiring which will then uh start the electronics integrating and classifying your uh your system so basically this is just a you know it says something similar you have the drive channel measured channel control channel which actually have pulses played on it this they're they're mixed up to a they're either mixed up to a microwave rf frequency and sent into the fridge and then they come back they're amplified and they are uh scheduled in the acquired channel where they're then digitized and stored as registers so base this is basically kind of what i like to show so on the left you have something called uh we call the circuit model of quantum computing and this is a hadamard gate on qubit zero and a c naught from qubit zero to one and what this does is generate the bell state zero zero plus one one so what happens if you actually take that and you see what's going on on the actual device level you get what's on the right and so there's a lot going on here basically every one of these channels is labeled a for acquired d for drive m for measure u for uh control uh and so your acquiring measurement pulses are all happening to the right here so we're going to forget about those and just concentrate on the pulse level i'm sorry the qubit control pulses so what happens here is you can see where the frame changes are happening uh these little these little uh circular arrows they are they are telling you where we're shifting the frame of the pulses so this will shift this will physically shift the phase of those pulses that are subsequent um and as alex said before if we keep those pulse changes aligned with the same kinds of cross resonance with the cross resonance channels then we just keep the we just keep uh keep accounting and we're able to do these frame changes uh naturally without worrying about it um so that's one of those tricks we do to reduce the error but then we also have these things called drag pulses which is derivative removal of adiabatic gate this this is a technique to try and reduce leakage on the pulses by adding a derivative of it to the quadrature or or imaginary channel we have an echo so this is kind of a standard echo where we do a c cross resonance one way we echo the uh the qubit and then do it the other way to try and get rid of d phasing and then we also have rotary echoes which will eliminate uh uh other kinds of interactions from the hamiltonian such as zz interactions which are undesirable for because you want to just get the zx out of the hamiltonian so as you can see there's a lot going on and if you look at the paper that's on the archive as last week for the quantum volume 64 you'll see even more techniques that were used at the pulse level to try and improve the quantum volume of those devices so i think that's very interesting and there's a number of applications i just want to kind of go through uh relatively quickly um like that i mentioned before so air mitigation is is one way we can we can try to amplify the error of the we we try to compile the best circuit we can for a given algorithm and then what we can do is we can amplify the error so if we just make the circuit longer we can make it more uh susceptible to incoherent error and so this this is the idea behind something we did um for chemistry called richardson's uh extraction uh returns extraction to the limit uh and what this mean what this means is essentially i have um i have an energy i'm trying to minimize right here but the best my circuit can do is with this amount of noise c1 well what i can do is i can add more noise to it and go out to a noise factor c2 and then i can measure my minimum energy on that level and then you can you can take the lambda it goes to zero limit and try to extrapolate to a new noise level so it turns out this this actually runs but you have to calibrate these these pulses up and then and then run them but it does produce better results for example this is from a variational quantum eigensolver simulation of lithium hydride this is energy versus bond length and the green the dotted green is the exact answer and the the white dots are the uh are the experimental results for this vqe we could only go to depth one so we couldn't generate too much entanglement because uh we had just too much error that accumulated so we didn't get uh very good results but but what happens when you run the same experiment but you go to a depth three well now for c equals one which is the best you can do at dep3 you can clearly see that's worse than uh than in the depth one case over here but now you can try and run it for uh more and more error and see what you get with the experiments and then by applying this extrapolation technique we're able to get to extrapolated dots that are much much closer to the uh to the actual exact answer and so this is a demonstration of how this actually works to you know per increase the performance of your quantum algorithm we also have different measurement levels of the measurement data and uh so yeah so since you were on that slide do you foresee uh do you foresee that ibm might um like so let's say we have a 10 24 shot and you know you're running separate experiments with this at a pulse level in order to extrapolate downward but what i'm wondering is would there ever be like a possibility to have that done um uh like uh uh like along some uh curve essentially uh for the shots themselves so if you were doing you know if if i were operating at a circuit level and i could like specify some option where um we're over like you know maybe 10 000 shots like that the additional error in the circuit is going to increase automatically for me and then there can be some extrapolation from that because right now i would guess that this is this is a lot of um manual like scheduling of multiple experiments and then you know classical post-processing right that's right and um i think you know there are ideas to do this because one of the one of the one of the things is we kind of have like a circuit model we call like quantum assembly which deals with the gates and then we have the pulse model which deals with the actual pulses and they're kind of like not really integrated right now but we have ways we can kind of flip between each other but i think in the long term we want to make these more realizable so that we can implement these techniques into uh like a more circuit based model and you can just like send a flag and say like do do the extra circuits for an error mitigation so i can't say when that's going to come up or if even but it seems like a very reasonable thing to happen again i have a question also i assume that basically calibration has been applied and they try to take into account and still you get this kind of uh off the actual value right and it means that you have to do the uh this mitigation process with different values of c right different values of c which i mean for this experiment they were calibrated separately um tisket pulse now has parametric pulses so you can you can possibly apply scaling um due to the intricacies of like microwave mixers and non-linearity uh so that you can arise there you probably won't get the best uh pulses in that case but you might get something good enough for like richardson's extrapolation to work okay thanks no problem okay so i'll just continue on yeah oh yeah can i ask you a quick question sure i'm curious you know how does this uh relate to your simulators uh do you try to uh run the simulator so that you've mimicked some of uh the profiling here or or does it work differently so when running the simulators i mean there's all kinds of simulators but typically what you'll do is you'll put in like depolarizing noise and and so that's kind of would work for this model and you can kind of get an estimate of how well your variational circuit will run based off those simulators which is probably good to do because vq is a pretty expensive algorithm all right great question thank you thank you good night could i ask a question as well of course you can um so so this actually goes uh to something maybe a couple of slides back where you showed this pulse diagram next to the circuit uh this one yeah um so it seemed to be what you were saying is is that these that these channels all uh kind of have have a fixed frequency uh and then this diagram is showing the amplitudes along at those frequencies is that is that this is pretty much true so i'd say uh i would say you'd have like at least for the control pulses you'd have three different frequencies going on right here or actually sorry two um yeah so for this one what we're doing is we are uh d0 is going to be at the frequency of q0 d1 is going to be at the frequency of q1 but u1 is going to be at the frequency of q0 because it's driving through qubit 1 because it's the cross resonance pulse these you know echoes these are all resonant as well but this is the yeah this is the one where you're applying it to you're playing one frequency to a qubit that's resonant different frequency but it's coupled to this qubit that's at the same frequency so for like a standard thing like this yeah they would be all the same frequency but that's something you can change also in pulse it doesn't have to be fixed frequencies so you can do like uh frequency shifts or or sweeps or whatever so is the is the the concept of a channel is it both a frequency and also a physical location that that that that pulse is getting applied to is that is that the idea yeah they can be they're probably going to be you know it's probably going to be a the same channel coming out of waveform generator and most practic most practical purposes it doesn't have to be they could be combined later but essentially the the these will be combined d1 and u1 will be combined physically and go to qubit one and then d0 and u0 are not pictured will be combined and physically go to qubit one um but the the two qubit interaction is the one that's happening between this one channel and this zero qubit uh-huh okay that makes i think that makes sense so great yeah uh it's it takes a little while to get your head around cross resonance so there's a lot of pulses and actually i've been having fun breaking these down and trying to figure out how to run them better myself because you can you can see how using shorter pulses might be able to get you a higher resolution or you'd be less susceptible to error so you might be able to improve your air budget if you could make these pulses shorter for example but these the the amplitude of these pulses is usually calibrated due to what's the shortest i can make them without getting too much leakage so it's kind of like the minimum you can do with a reasonable error rate and you and you get leakage i guess because you're applying too much energy that's exactly right on that channel is that that's exactly right yeah and so those are the kinds of things we used to calibrate and in fact it's it's it's shown in the uh in the paper that that's on the archive from last week you can see those those exact curves we call them tick plots about how you how you choose what length to make these cross resonance drives at but yeah so this is all stuff you can play with yourself as well um and that's in the volume 64 paper that's all that's in the volume 64 paper that's absolutely right uh let's see other applications besides error mitigation but this is you know an important one this is actually you know a classical technique um there's certain kinds of you know waveform level data you can get back from your measurement for example so so right now there's too much data for us to transfer we're still working on this but we call it level zero measurement it's like a raw waveform that you get mixed down uh to to an if frequency of tens to hundreds of megahertz or something um so typically what you happen with that that data what happens on the computer is you integrate that waveform with some sort of kernel so we call it kerneled data which is level one and you integrate it to essentially average over some point in the complex plane or we call the iq plane which corresponds to your measurement and so that's an iq point and if you're measuring zeros and ones you tend to get these blobs or clusters of of the eye and the q and then you can imagine well uh i want to tell the difference between zero and one so i discriminate in some way so in this in this picture just i'm showing you you can draw a green line and say everything on the left is is a is a one and every or sorry everything on the left is a zero everything on the right is a one um but you could be much much fancier you can draw other kinds of things you can build distance of gaussians you can do machine learning you can do all kinds of things and in fact with the um the the excited state uh sorry second level excited second excited level state readout that opens you up to a lot more classification possibilities because you have more blobs to deal with you have you'll have three instead of two so you'll we'll see how how this goes but this is very interesting to how how do you do and that i gotta say that improve the measurement failure quite a bit just being able to classify these things better so it's not always just like the physics just to show you kind of what what you can do with the measurements is you can do a small measurement where you don't get much separation between the zero and the one states and as you increase the measurement length and you integrate to some degree you're going to pick up more and more information from that measurement you know to the point where you can relatively well classify them and it doesn't get any better and in fact due to relaxation it'll get worse uh so so you can have you can play with these things and try and optimize these things and this is even messing with the pulse shape of the measurement for example which is something we've done in the lab as well uh which you could also do with open pulse we can look at the high order levels of the transmonds these are useful for other things not just readout so we've got the two right there there's some applications so we can we can measure the the energy difference between that between those two states is something we call the anharmonicity it's related to a device parameter called the charging energy and that's just good to know from a device physics point of view but you can think of other gates that use that higher order level or you can make a q-trit or something like that um so there's some ideas of having a to follow gate using the higher order level essentially by doing like two cross resonance on the on either end and across resonance in the middle between the one and the two uh other things you can do is you can you can excite your state into high order population so you can get some really good contrast to measure the temperature the effect of temperature of your qubits um so the residual excited state population is discovered by this fermi function here so you can extract the temperature and then i mentioned uh using optimal control you can do leakage reduction so this is a thing instead of using the normal drag pulse which is in the in the dashed line the this group used uh piecewise constant functions and optimal control in order to make a four nanosecond a single cubic gate with higher fidelity than we have typically in the lab so you can improve a lot of things using a lot of these techniques so i guess in general i uh brings me nicely to my conclusions is that you know this pulse level access allows you to squeeze the most other performance and in fact like i think the the qv64 paper from last week really shows you what how much this control has an effect on the ability to do things or to do computations well but of course you know it's both hardware and software that we need to focus on and so you know we're in the lab designing new hardware fabricating new hardware building better fridges but we also uh need to come up with more ways we can control things in different ways of uh of you know optimizing for that control and then coming up with different metrics that you know device level metrics usually tell us about one one or a few things but the overall overall system metric is something we're also continuing to uh push forward so uh i'd like to thank you very much for your time and if there's any remaining questions i'm here yeah i've got one for you sure what are you uh most excited about that you haven't been able to uh dig into yet in terms of um i'm sure you've got a lot of experiments you'd like to try i'm just curious like knowing even knowing that like not all of them will pan out i'm just wondering what where your intuition is about what uh yeah what you're excited about doing next so i think um tailoring these cross resonant pulses to to more use cases is something i'm thinking about but i'm already like i'm not going to say i can't do it yet because i'm trying to do it right now other things are a more level of control and feedback is something that i think we'll see progress on you know pretty soon but it's it's a little hard you'd like to have like an if-then statement in your quantum program but right now you have essentially all your control uh first and then all your measurements afterwards but you know this is something people are working on uh so that i think that's probably the i've been waiting for that for a long time too so that's that's tough um can you kind of explain again what was the main uh component that increased the quantum volume to 64. oh absolutely so i didn't put in i only just made this one slide for it for it i didn't go into the details of it but i hope you know we'll have maybe on our um our kids kit youtube channel uh a talk about the quantum volume 64 paper soon um but it was essentially all all of these uh things so essentially you're trying to do so the first thing is improving the compiler which i didn't talk about uh basically quantum volume is judging how many how what's the size of a random uh a random quantum algorithm i can do a square sized algorithm where each component is a is a qubit or a random unitary uh between two qubits or between any qubits i guess then you you can you have to map it to the ones that are actually uh there um your actual architecture so the thing that was done in this compiler was figure out better ways to map those uh random circuits onto the qubits that actually exist so for example our superconducting cubes don't have all the all connectivity now right now they're what this is ran on this has something called a heavy hex connectivity which means we have um like a it looks like hexagons with qubits between each point on that hexagon so like each edge has an extra qubit um so you can't you have to figure out how to map something where there's arbitrary connectivity to something that has limited connectivity and that's that's essentially one thing that was improved with the compiler uh this excited state readout it it promoted the um the assignment zero and one uh became much better so that doesn't just help your measurements but that also helps um helps your calibrations because you can get more out of your calibrations if you have less error rates uh the shorter two qubits gates this is probably this is probably one of the hardest things to do we've been getting better and better at it but the two cubits gates seem to tend to be where most of the error happens um and by making this shorter we were able to you know i think we went from 280 nanoseconds to 200 nanoseconds for your uh your average two cubic gate it's all it's all in the paper i'm just trying to pull it from my head oh yeah and by the kiss kit compiler they were able to cut down the number of c knots for from like 9.4 c knots which is the 2 cubic gate 9.4 c naught per unit of depth to 7.3 using the compiler so you save a lot of c knots and then saving 80 nanoseconds off the c notes may not sound like much but it is a lot when you have a lot of them and that's dynamical decoupling scheme is what we realized and actually we've we published a paper on the archive recently about the rotary echo technique that we're using and essentially what this does is if you keep all the spectators busy they have less time to mess with your operations that's what i'm going to list things loose framing of the situation but if you read the paper you'll find that it took all these things to get up to quantum volume 64. none not one of these none no single one of these is what did it uh but these are all control things that the important part is you needed all these four things but they're all control things they're not hardware things thanks no problem hi thank you very much for a great presentation i have a question on uh in general on the superconductor cubits i think the qubit is operating under very low temperatures like a few kelvins that's right and there are higher super uh higher temperature superconductors why are we are not using the higher high temperature superconductors but instead it's very low temperature wise yeah uh that's a great question um a lot of it has to do with the the high temperature superconductors are um very difficult to manipulate very difficult to fabricate they're usually made out of their like cuprates for example which is like your coffee cup um so you can imagine how hard it is to kind of fabricate that and and and turn that into something especially growing like a oxide layer between them whereas metallic ones that are typically used these days are uh are going to be a little bit more malleable they're a little bit easier to process and things like that so from what i've heard from people that are experts in fabrication it's really hard to get a uniform uh oxide layer between these things and they're really hard to manipulate uh to begin with uh that being said there are people that are working on this this is definitely an area of active research you know essentially everyone makes aluminum aluminum oxide junctions because it's easy to make them and we can take that pretty far but there's also a lot of people um in you know especially in universities that are that have labs that are doing much more you know highly controlled growth techniques in which to make more pristine junctions which could maybe eventually include high temperature superconductors i see thank you thank you very much no problem any other questions uh if not let's thank nick again thank you nick that was awesome welcome thank you thank you for your time uh before we go to the second talk with andre i just want to ask again if anybody got any announcements to make uh looks like not so then we welcome andra koenig uh a man of many hats i think uh you have uh your affiliation as interference advisor and uh other so please you know introduce yourself and tell us more and uh we're eager to learn uh where the uh money goes in the quantum landscape absolutely uh thank you alexia and thank you nick for uh that great introduction or refresher for for some and congratulations on the quantum volume i know ibm is committed to doubling that every now and then so so looking forward to 128 soon um i'm a data kind of guy and i do a quick introduction uh this gentleman that many of you probably won't know edward stanning he comes out of the manufacturing sphere and quality around manufacturing said this very famous sentence if you're in this kind of thing um without data you're just another person with an opinion and that is kind of the underpinning what i do we'll be looking at a few slides but very quickly move into some of the actual data myself i do wear a lot of different hats um i mentioned earlier i was a management comes out in four dozen years with the big houses um starting with anderson consulting in the late 90s accenture ssa and company and so forth you see some of the customers i worked with many of them at the senior executive and c-suite level i went to a bunch of schools to study mostly economics and business the first one university uh silence a german university where i'm from um studied economics finished the degree at the business school icn in france then got an mba at the the university of chicago um and unfortunately i'm not smart enough to uh ever dream about getting a phd from mit so i just took all the quantum computing classes to to at least get a little bit of sense and bought a big golden frame for for my certificate to show off a little bit um if i don't do data economics and all these things you'll find me on a sailboat one of the reasons i live down here in miami here you see me competing in a national championship five years ago today mostly cruising through the caribbean drink in hand i run three companies all of them within quantum technologies interference advisors is kind of the gardener of quantum technology so we collect data sets produce insights and write market research reports we've been doing that for uh over three years now and the leader in the space um one quantum the second company is a community organization organizing chapters we have a startup chapter currently with over 30 members and growing that we only launched a couple months ago a woman in quantum chapter where we had a big conference a couple weeks ago to kick it off with 652 attendees we reached over 150 000 people online had 16 speakers over three days of content so trying to build communities around themes around geographies will hopefully be starting quantum africa soon since i have a small team in africa to build community from anybody in academia business entrepreneurship vendors government investors that works within quantum lastly i'm also an investor through my fund entitlement capital um where i'm currently working on two deals a three million dollar round for a quantum tech specific startup accelerator um kind of the follow-on program to the destructive construction lab in toronto that was mentioned earlier as well as a 60 million euro quantum computing hardware deal that we're currently trying to put together for a european company so those are my three hats that i wear everything that i do going back to damnings is motivated by this idea of data and you know this is a very difficult topic but i'll always remember one of my professors from chicago richard thaler who introduced us to this cognitive bias codex and there is research the better you educate the more professional and life experience you have the more prone you actually are to having these cognitive biases things like an anchor anchoring bias where you know when asked for a specific number and somebody said 314 startups when we did a little pull earlier it's probably anchored to a number that they previously in a different context um so you see this codex there are many different types of biases uh heuristics and anchors that we unconsciously use that inform our making i certainly do that on a daily basis so really hoping to bring data insights facts analysis to try to mitigate that as much as possible and really mostly for the investor space that that is my main target and that is what we'll be looking at a lot here during during the next few minutes um we really see that yeah a lot of the investors do not understand the purpose and the sense of what quantum can do as a technology they also don't understand the market and how to scale it on a revenue basis for specific exit or different exit scenarios that might not be based on a revenue type model and the technical details in matter but also the data matters because there's a lot of hype there are a lot of people out there that are very light loud we don't just have noisy intermediate quantum computers we also have noisy intermediate quantum startup founders and we need some some data to fact check them um so the idea really is you know to collect that data not just provide tables and you know articles and news but really try to understand what insights we can gather out of this data and you know hopefully turn into wisdom for our clients which are investors but also regional governments um in north america and europe as well as vendors and some of the large startups the way we do that is we actually have a methodology and technology to continuously collect data um there are two things that we don't do the first one is to simply write down any kind of data that we hear you know i'm sure all of you hear about startups or or qubit count or cubit volume there's a lot of blocks out there that kind of collect that that is not how we approach it we actually systematically parse you know public as well as private um sources of information we have methodologies to go through social media channels private networks but also conference conferences to mythologically uh collect this data the second thing that we don't do is consider data that is not verifiable so everything that you'll be seeing in a couple minutes here is verifiable data verifiable i mean it's data that you know can be proven publicly because there's a website something on linkedin a press release an investor that backed it or a private source that is able to verify it what we do not consider is again hearsay opinions claims assumptions and so forth um we really try to stick to the facts then obviously we're not perfect in that but uh getting better every day is the ambition and goal here we then look for insights and how do you look at insights it really has to do with looking for extremes for tail ends for spikes for anomalies um looking for changes over time looking for relationships looking for gaps so we actually have specific frameworks um for for much of it yeah some of it we do by by instinct or habit still unfortunately but a lot of it we we have you know spreadsheets where we look for certain uh of these potential insights we score them according to to a system um and then we see what comes out at the end so so methodological um effort to look uh to derive and arrive at these insights ultimately the last step and i'll show you one example is to use all of this data and the insights but also our networks that we have to produce market research reports similar to what our gartner pitch book crunch base would do to inform the decision makers decision making of investors government officials enterprises clients and so forth the last important step in um achieving that is the importance of importance of data visualization and i mentioned that not just showing tables or or simple shards and this is really something that we're pushing every day and we still have a lot to learn and a lot to improve but uh making the data as interactive intuitive personalized and you know collaborative as possible so that we can do a better job at arriving at insights and our clients get a lot more value out of that data by by interacting with it and a very simple example this is not quantum related just a typical hr table like like you would see it in an excel spreadsheet certainly something i've done for many of my startups and then an example of something you know color coded filterable searchable sortable that is a lot more visual and insightful and this is obviously just one example of what you can do but the importance of data visualization in arriving at insights and meaning is really key and something we focus a lot on um that concludes the slide version this deck contains um many more slides specifically geared up investors and how investors can make better decisions within quantum tech and and uh startup ecosystem um here we propose a segmentation framework for the entire quantum system um how to benchmark quantum tech deals which is very difficult because there really is no precedence it's a young industry how do you get a quantum tech pipeline uh deal pipeline as an investor and how do you evaluate it how do you ultimately make your investment decision and successfully operate your uh your your investment um so we propose a lot of frameworks here we'll be sharing these slides um if anyone is interested in that but let's go look at some of the data get this out of the way all right get rid of this there we go so we asked this question earlier how many uh startups there are in quantum technology and again looking at the whole quantum information science space across quantum computing hardware software but also sensing communications um quantum key distribution and other applications we have a little chart here that shows how this has evolved over the last three years um uh and you know a couple interesting um takeaways here in my opinion number one the space has grown significantly since 2018 when we first started tracking it but in 2018 already it was you know a very significant space um i don't know if it was 314 total i'd have to check but in 2018 um there already was about three to 350 startups globally that we were able to verify and track as you can see and i'm highlighting quantum computing hardware here this has grown um but it hasn't exploded over the last three years it is something that has naturally grown and evolved as the technology and the market has matured um but really took uh you know it's uh it's the first steps um already years ago the other interesting and in my opinion um inside here is that we have a very large number of quantum um software um algorithm and simulation companies sometimes it's hard to distinguish um amongst all the marketing talk what the company really does do they just you know do algorithmic research do they actually write software um do they ultimately do simulation but there's a very large ecosystem of let's call it quantum software startups and we have a separate category for only quantum computing software startups here so if you combine these two there's more software out there than hardware which i find very interesting considering that it is an industry that is still young a technology that still needs um a lot of development and breakthroughs um if you compare this to classical computing i wouldn't be able to say how many software companies there were 50 60 years ago but but certainly um far less than than hardware i i would assume so this is a very interesting uh finding if you look at some of these other categories you see that overall they are much smaller you see that the growth is much lower and you also see that there is more and more folks like me um and that might be a good or bad thing uh consultants think tanks um today we actually count uh almost uh 40 of them worldwide that specifically do consulting and a lot of hard thinking about the business side worldwide about quantum computing a lot of media outlets that have uh sprouted up and then some other companies that you know do quantum learning and other kind of aspects of uh quantum technology so this is kind of the ecosystem um unfortunately it is very difficult to get data out of china and russia at least verifiable data so you know take that with a grain of salt if you look as of june 2020 at the geographical distribution of it no big surprises here obviously north america and canada the u.s leading here in terms of the number of startups 122 in the u.s 43 in canada europe and this is surprising to many uh in aggregate has northern north america and individual countries such as france germany italy and others are growing very strongly the uk if you look um you know even so they left the european union um other than the european context we currently count 64 quantum tech startups so a very strong um and you know strongly growing uh network of companies there um very little activity unfortunately in latin america um i'll mention our friends at quantum files with martin and rafael um you know three companies in south africa a very good university there we then do have some activity in asia with india and china we have only one truly verifiable uh startup in russia obviously there's a lot more than it and i think everybody knows that australia is very strong um the startup ecosystem so still relatively small um japan i'll be very transparent we haven't done a great job at collecting data yet nonetheless we already count 17 startups to date within quantum computing we also have the breakdowns by applications of anybody who's interested in you know quantum computing versus sensing communications encryption software or hardware um we can look at that separately so that's to get us started some of the startup ecosystem next we'll look um at the market but wanted to see if there any questions at this point uh andre i just have a question so um so basically the the coloring are upper brackets right so uh 20 means up to 20. so when you look at south africa it's three but so it's colored that 20 colors it's under correct and i you know we can um change the granularity um good point actually maybe maybe an opportunity here to make it even more granular in in five increments or whatever but that is right yeah and one more question so uh obviously there's been a few years and you know half life of startups sometimes is under two years do you track this historically can you move in time and see which startups were born how many died you know in a given time period great question and that's exactly how you get into insights right tracking this over time looking in what regions certain companies are evolving are they getting a specific type of capital do they work in a specific qis application or are they dying and we have what we call a graveyard so we do have a whole table and shards of quantum tech companies that are dying um there aren't too many at this point and and we haven't seen any trends in it yet um we have not so focused on that piece of analysis yet um something that's on my to-do list um my my opinion if i if i can have an opinion to to to say something against what i just presented is that there's no insight in what has died or not survived so far we definitely see very interesting trends as to where hardware or software is stronger in certain regions um and how you know certain regions might do more quantum communications versus quantum computing and those types of things so that's that's surely um available interesting uh you mentioned the startup one revival started up in russia is it by the chance called curate or something um i'd have to go into my database the name doesn't ring a bell right now i can check that after this call uh the curate yeah these are folks from uh skodak which is the innovation hub uh near moscow and actually they uh they were in our conversations and offered to to speak about you know the work so i'm just kind of making a connection here uh there they are an interesting company yeah so cool well let me see this so i have a small team of folks that help me collect clean analyze the data write and produce insights and reports i would love to find somebody in russia and somebody in china who can do that for the local market so if anybody has has referrals here um send them my way we'd love to explore those markets and get great data absolutely i'll do so for russia and if folks can do it for china with you know awesome thank you um also a question and we're slowly moving you know towards um on the investment side of things uh but just looking at the global market right and there have been a few exercises out there in in terms of sizing and forecasting the quantum information science market um many of you might have seen the boston consulting um a group effort from 18 months ago obviously tremendous piece of work um there are other you know market researchers out there that estimate the market we build a model which i'll show you in a second here um to really do custom market size forecasting for quantum information science and what we're forecasting here is the revenue potential for quantum information science vendors which might include you know our friends at ibm and their competition startups and so forth in commercializing their products and solutions there's another type of market size forecast that we could do and hopefully we'll publish soon which is the impact that quantum technologies will have um we mentioned jpmc a few times uh during these conversations already jp morgan obviously has a huge program the impact that quantum technologies could have a jp morgan other banks chemistry and so forth far exceeds the revenue forecast potential and and i think that would be very interesting to look at if we look just at the market size forecast for revenue and this is a very conservative scenario you'll understand in a second what i mean by that we estimated at 172 million us dollars for this year we estimate that by 2030 annual revenue potential will be around 2 billion dollars and a compound 10-year market size in terms of revenue of 9 billion dollars you see a red and green box underneath where we do a little bit of scenario planning so this is what we call the base scenario if we look at the low scenario um those numbers go down um year one 2020 doesn't change but the 2030 obviously a lot smaller and those are the compound growth rates kicking in here and a total market size that is only a third if we slightly um improve our assumptions and make them slightly more optimistic um there's really an explosion in in the market size potential year one 2020 obviously doesn't change again but you see that in 2030 um um you know we have a slightly larger um market the compound however is really tremendously um more significant and and that is the real insight here we'll look at some of those assumptions in a second here for this base scenario we also have a breakdown by qis um application quantum computing no surprise makes up the bulk of um you know that forecast uh quantum encryption um already is a market today they're already companies that make significant revenue to the tune of 10 20 30 million a year quantum sensing quantum communications and other things you know really only take a small share let me actually show you and this is something that very few people see the model behind it um there there is a very large uh let me move my speaker thingy here a very large flexible model that is based on a number of variables and assumptions by number of customers average price market size and so forth what we do on this model is that we segment it in applications quantum for finance quantum for chemistry quantum for logistics machine learning new materials on encryption and so forth so we have all the potential use cases that quantum could deliver because that's where the revenue comes from right the revenue doesn't come from a quantum computer itself the revenue comes from jp morgan buying cloud credits or service level agreement from ibm um or eventually an on-premise quantum computer from from somebody else um so that is how the model is is structured um we then estimate hardware on-premise sales and there is some in 2020 which which surprises many people we look at hardware service level agreements which is something that you know large vendors tend to do where they give you a one year two year three year contract for access to their hardware which also includes software which also includes services um joint research projects and and these tend to be fairly large service level agreements there's a lot of we looked at that just a few shorts ago software companies out there as well as simulation companies they tend to sell smaller product solutions platforms maybe for 50k 100k 200 000 a year and then you have hardware cloud right we've we've all been uh on ibm um amazon bracket just launched righty where you all have you know you have a free layer you tend to have a free layer that seems to be changing for some of these vendors but then you have to swipe your credit card um to continue using it so here you can estimate the number of customers down here you can estimate the average price point for each of these and you can estimate the growth rate uh the conservative base scenario that i showed to you earlier in here is uh based on the smallest assumptions so you know really starting with one customer in 2020 per all of these segments um the lowest price which is as an example 15 million for an on-premise hardware computer i i believe there will be three or four sold this year one has been sold already um and and the prices were around 20 20 12 million um so these kinds of um very conservative assumptions and then a growth rate for all 10 years um which i had put up um let me check uh the 30 uh which which i believe to be very conservative and low scenario the growth rate is 10 and the high scenario the growth rate is 50 so this is how this model works and this is how we estimate the market this is also how we work with customers who might be interested in predicting forecasting a specific segment of the market quantum machine learning or quantum computing cloud and then we can very easily and reliably do that with the scenario planning tool question andre okay great question if i make do you have any data uh before with you on quantum annealing for people versus universal quantity computers in the market because this is the difference between the market these two paradigms are in fact aggressive and if you have a view this yep great great question we do look at the d-wave and such um in some of the investment um data that we'll be getting to next um and for very specific reasons that that will explain uh so yes we do that in this market ecosystem where we started and market size forecast um approach that we're looking at right now we do not include annealing um obviously it's a big debate right it's a healing quantum or not i think it's it's a little bit of of a silly debate to be honest because at the end of the day um d-wave has a lot of customers and really delivers value for them and and that's what matters but that said we did not include it um in the ecosystem slides you you've seen so far or in this forecast model um i also don't have reliable revenue predictions for annealing no thank you good so we've looked up giving it away here we've we've looked at a little bit of the ecosystem we looked at how we think about the market and how the market might evolve um so by the way i didn't mention that we have 542 quantum tech startups in our database so far so we're in between the 100 and 1000 that all of you estimated earlier um to give you a little bit more for a sense that grew by you know i'll say about 75 over the last three months alone um so yeah um a significant growth rate um for one part another part that we are getting better and better at finding these companies verifying them and including them in our data set so it is an ecosystem that is growing strongly um it's not exploding as we showed in the first chart the bar chart um and it is a very large ecosystem but we're not at the 1000 startup mark uh yet i also asked you to estimate the usd investment amount for 2020 so far and here it is all the verifiable investment deals for the first and second quarter of 2020 386.5 million u.s dollars in private capital these are not grants these are not government programs but private investors that have invested into any of these quantum information science um and this to your question just a minute ago would include annealing and d-wave if they get any money is the overall amount invested so far and this is a very large amount if we look at the trends over the last three years a little smaller um the overall fundraising amount um that is verifiable again we're only looking at verifiable data right i'm sure there there are rounds out there and fundraising that that haven't really been announced that you know might be rumored but aren't verifiable we don't include them all in all you see that total investments on a yearly basis you know somewhere around the six seven eight hundred million um a year um 2020 in the first year we are you know close to 400 million and that during the biggest crisis that any of us has ever been through um and where many other sectors of the economy really suffered in terms of fundraising um so to me this this is a great sign i don't know who still talks about quantum winter but from a financial point of view uh the first two quarters of 2020 clearly have proven that there is no quantum winter and very significant private funds flowing into the market you also see the distribution here by qis application quantum hardware is getting the bulk of it um quantum software and qkd we haven't seen much in communication sensing or other fields at this point if we look at and this is interesting at where their investors come from so these are not the target companies but the country of origin of the investor no surprise america the usa um this is the number one country here um and that is going to be a long time before it changes this is the number of deals so 26 deals made by american private investors but interesting here is a couple things you see countries like india starting to invest in japan you see activity and then when you look at europe and again europe and aggregate five deals um by french investor three deals by german investors uh finland um norway uh and then again the uk with ten deals um we're kind of exceeding um you know american activity uh in terms of where the capital comes from and this is very very interesting um if you're european and german like me and a founder entrepreneur it is always this complaint and justifiable complaint that it is so hard to raise uh capital as a european entrepreneur in europe and that certainly has been true and digital in sas and artificial intelligence blockchain it is not true in quantum and and that's a big surprise um european investors are um you know very very excited about quantum and they are opening their purse strings um something that is not on the map here but i'll mention it um israel is something that everybody needs to watch um a lot of stuff is happening in israel and that's going to be a big fat dark blue spot on on the next edition of this map any questions comments we'll look um at the next chart which looks at the target countries so where these deals are being made and that is equally interesting uh obviously again the us is number one here but only with 14 deals so we had 20 fixed um investors or investments out of the us but only 14 deals being made in the us and that only leaves us with europe right and really a lot of deals currently happening in europe we have some in australia canada is always on the map and always will be but here also we see that the deals are being made in europe increasingly and even foreign capital uh is moving into europe which is unusual at such an early stage of a technology um so american and other foreign investors investing into european hardware software qkd companies so far so good i have a question go ahead so so looking at at this map and then the previous one there seems to be more sources than there are targets could you explain what's going on absolutely um if if you're a company you're going to have many different investors right all right we have only one investor you typically have what's called a lead investor um say you raise a 10 million round that lead investor is going to take three four five six seven million and then find co-investors um other people that fill up that round so you know an investment round can have anywhere from one to five to a dozen investors um you're definitely going to see five to ten um on average so that explains that that gap all right let's look at some of our other data um top 10 quantum tech startups by funding so far um everybody read about psy quantum uh hopefully that is no surprise to uh to anyone uh this year um really really amazing with half a billion dollars uh this goes back to the question earlier annealing um d-wave still in a solid second position but i i'll have a remark to that at the very end of my presentation in a few minutes here um regatta you just announced uh uh roughly 80 million by a very large new um professional venture capital fund bessemer but also other co-investors so sorry getty really progressing quickly uh our friends out of cambridge um continuing to do very well you notice id quantity so quantum encryption um but this gives you a sense of you know how much um certain companies raise which is very impressive but for somebody like me who has worked in sas and software artificial intelligence before these are fairly small numbers and the fall off if you start to look at the you know iron q and then id quantity um these numbers for top 10 list get them relatively small very quickly and then you see xanadu and 10th place here with 35 million and then you know very long very flat tail if you look at the rest of that list um so so there's still a lot of room for improvement um yeah that concludes the investment part i have uh one more investment kind of finding that that i want to share in a little bit i'll show you another a couple slides any questions before we move on all right so i'll show you some of the other things that we track um you know obviously we just had a great presentation by nick um ibm is the king of the hill when it uh comes to quantum computing but um quantum computing performance and access to quantum computing um so here you see a little distribution of qubit counts also by modality we have maybe more interestingly i don't know if that's the next chart it is um quantum computing cloud access this short surprise is many there's a lot more um quantum computing cloud out there than many think um everybody is familiar with our friends at ibmq we all know that righty um obviously has access um you know amazon is out now with bracket um so so you know nothing new there but there's a couple dozen other very small offers that come out of you know overseas vendors universities and so forth that offer uh quantum computing cloud access to mostly their own hardware and then you have some reseller relationships such as cqc um who provides access for preferential terms to the honeywell hardware for example um that are part of that list as well but this is kind of um the whole quantum computing cloud access market with a total of 29 players to date we then also for our clients start to you know dig down into some of the applications here's an example of quantum chemistry startups and we currently count 20 that you know specifically state us their mission to do quantum chemistry um you know obviously there are many others that also do quantum chemistry amongst many other things but this is just an example of how we try to look and segment the market here we have a short where we look at a head count and this is also interesting to track you know the evolution over time comparing this you know to the previous six months a year ago and so forth of how companies are evolving uh very difficult to find data here right so this this is to be taken with a grain of salt but if you look at it consistently over time it does tell you interesting things so we go by linkedin and look at headcount to me it was surprising to to learn that uh one cubit as the largest company here um you see folks such as strange works um um a number nine spot um great company but you know to me it was surprising to see them so so far um a lot of these startups are still very small that that is one of the key takeaways of this year um another second key takeaway is that there are other startups that i know are bigger that didn't make the top 10 list but they don't report those numbers and um i spent a lot of my time encouraging some of these founders um to to do a better job at that frankly at the storytelling at their messaging at their information sharing because this is the kind of information that folks like us look at you know customers um government investors and so forth and if you don't share some of that information you just make it you know it's an additional complication to overcome beyond understanding your qubits and your error correction or whatever else in terms of assessing the viability of you as a provider potential partner or or investment target um so those those are the two key takeaways here but all in all headcount is still fairly small um could i could i ask why qsoft is in the list there i thought this was a this was an academic department only a good good question sometimes it gets very very difficult to um and and that's done my own freedom of of decision making um uh to decide if somebody is as a company or not um qsoft has a lot of commercial aspects um they have a legal commercial legal entity they have commercial offerings um it is very hard to tell if they actually are real company or not um in this instance i decided that yes but but many times that is very unclear and still you know hard to tell so there's a little bit of a judgment call here but qsup is commercially very active um one more oh and okay so sorry uh this actually was just the top ten quantum software startups by employees um i i reversed the slides here the top 10 quantum tech startups overall um um that's the chart here um so that picture looks uh looks a little different a little higher um head count all by almighty not by much we see some more bigger and familiar names um we still see a considerable fall-off in in-head count for top 10 list here which which indicates that there is a lot of room for growth uh left no cue soft on on this so that's uh that's good news these these are definitely all commercial companies um uh but again you know um information is difficult difficult to get in the sense but you compare something like this to funding to come you know number of publications which is something that we track we track patents by companies and so forth um so these are all small puzzle pieces and an overall picture um that we're trying to paint um and in that sense it is you know one very important insights um but obviously not a deciding factor um all right something i just just wanted to interject uh you know i've been in a lot of software startups so it's interesting you know these numbers may look small to somebody who worked for ibm for a long time or like is used to you know big companies but in kind of bay area these are all huge startups right because you started like with a couple founders if you grow to 50 it often means you have at least 25 million investment right from a top vc firm salt so like this is all to me they look like awesome like you know 200 obviously it's it's it's you know it's near unicorn like you have companies like data bricks with multi-billion dollar valuation software with 300 people so just to put it in context you know this all looks very healthy and promising to me oh you you're absolutely right thank you for bringing that up you know as soon as you have more than 40 50 60 employees as a startup um you start to be a very very serious startup that at least has raised you know 50 100 million and so forth um absolutely a very very good point um so two more things that i want to show to you and then we're done here um um we have different databases so far i've shown you uh you know data and insights out of two databases um the first one which we call our entity database and entities can be a startup a research lab and we have you know the same kind of insights and visualizations for research labs um could be government entities and so forth so that's our entity database and i'll show you some examples there we have an investment database and that's where we keep track of all the dollar amounts the deals being made who the lead investor is who the co-investor is and so forth we have two more databases and i'll show you uh in exclusive preview some data from one of our third databases which is the quantum computing use case database so uh we track our use cases again that are verifiable that you either have been published reported or or been verified um um in private by you know the key stakeholders we track what the use case is who the vendor is who the end user is what industry what country it is in um so this is a very interesting list and when you start to dig down into this a little bit um we talked about finance and banks earlier and i think alex you made an observation about that you will very quickly notice and here's an example on page 3 that financial services banking insurance is absolutely the dominating sector in this use case kind of research that we do so just wanted to share this with with anyone with everybody um but this is kind of another point of view that we take at the industry all right um everything that i showed to you right now and a lot more is available for free with a caveat on our kiss data portal qisdata.com so you can go here and find a lot of these charts for free what you will notice is that we typically give last year's or previous year's charts away for free um and you can simply download them from here um if uh if you want some of the um up-to-date current charts uh custom charts or more highly segmented charts um then that is for our paying members uh where we have kind of a subscription portal to access that um so any of these free shards feel free to use them if you do your own sales pitches investor pitches whatever if there is a paying chart that you're really interested in that you don't see here for free um just send me a message and and we'll figure out how to get that to you uh lastly we use all this research and all this data to ultimately write reports um here's a report that just came out this week by one of my senior analysts foreign which looks at investment trend analysis in quantum information science um and here uh we get back to the d wave question and uh we'll uh we'll end on that um one of the big insights is that if you look at the data over the last three years um let me make this a little bigger nope can't right now um if you look at the data over the last three years um investments have very aggressively shifted away from annealing um into superconducting but also trapped iron and photonics um and annealing seems to have reached you know i'm not going to say the end but the peak of its fundraising journey and really more of a reliance on revenue if if they can succeed with that but maybe we'll be surprised by another big d wave fundraising deal but really the main activity over the last three years and that is something that started over a year ago has shifted into um a real quantum computing modality so we have the breakdown and the explanations of that and here if anyone is interested let me know but these are the types of reports that read and crews produce for vendors investors governments and so forth that concludes the presentation the slides that i shared earlier are available on my website simply go to undercunic.com um you'll find a link at the bottom of it you can also email message me on there if you're interested in any of the shorts and i'll be happy to have thank you very much andre excellent overview uh any questions i was just wondering who bought the quantum computer i can't tell you but i'll i'll say this um there is a big pipeline of regional research labs that are looking to buy on-premise quantum computers um that don't want to build them themselves or can't build them themselves um you see that in europe you see that in asia um i don't know of any end customer yet you tell me if you guys have sold any on-premise i don't think you have but you know there's no bank or something like that that i'm aware of um but there's at least one european research institute uh that has bought one and uh i know of two or three others that um yeah are on the line to to cut a deal by the end of this year interesting yeah it seems like uh it's something hard to get out of the lab well you you can you can buy it on ibm.com but exactly you put it on the cloud you don't need to have it in the lab i mean you want to you want to go maintain that stuff not not everybody wants to be a dependent um you know of an american cloud provider fair enough i i remember that a few days a few years ago actually nasa bought a d-wave computer for from them yeah 2000 2000 absolutely the d-wave sold several as as far as i know which is not counted and the data i showed to you but yeah now there there is some there are research labs that say we want our own on-premise a quantum computer so we can you know really not not sit in a queue but really also experiment with the technology upgraded to our needs or over time and listen there there are folks out there this this is a little bit about geopolitics that say we don't want to be dependent on the cloud number one um number two we don't want to be dependent on an american member or a chinese vendor or russian vendor whatever it might be right especially if you're if you're a european country and this could be a transformative technology um so so they have different buying behaviors as well i think ultimately in 10 15 years the the quantum cloud is what everybody will use right in the next couple few years i think there will be a big on premise market yeah uh so andre i have a question about um so it's obvious how investors would use this data right like you would look at the trends you would look at the deals make judgments right and see what other opportunities if you are a researcher an engineer maybe somebody thinking about a startup all right so how would you use it from the other side for instance let's say you're young uh researcher and you decide should you stay in academia or should you go to ibm research or should you join one of these guys or should you start your own startup how would you use it to kind of make judgments and support your intuition yeah great great question um and and two answers to that i've i've been on a lot of meetings and calls and meetups like this and many times i see engineers developers using you know a couple ecosystem slides to to frame the discussion of of what they are doing i think that's fantastic um many are familiar with the quantum computing report doc think who's kind of pioneer of that and does amazing work um so i i offer an alternative for that and maybe a little bit more variety i'm not trying to compete with that too much on that it's very different um but but that would be one aspect right and if anybody for their own presentations need something specific please reach out to me um the second answer and i had that conversation a couple of times this week with researchers asking me how they could transition to entrepreneurship um i i think there tends to be you know a lack of really full understanding or maybe some bias because there is a lot of hype in in some news publications you know somebody who said earlier i think um investments total 2 billion dollars which which unfortunately is not the case yet um so getting a better sense of which type of quantum information science fields applications but also regions are maturing faster sooner or maybe later that might be an opportunity as well right depending on your appetite and your skill set but just getting a better understanding of um how how users um um enterprises and investors evaluate quantum which is very different than a scientist a researcher would evaluate it i think it's very helpful if you're trying to join a company as an engineer or potentially start your own um so i think it's helpful from that point of view yeah uh and one more question so obviously all these startups uh all uh all of them need need uh engineers uh so you know like in in software startups you have a thriving jobs like a system recruiters headhunters right like it's a war basically right like if you're a machine learning guy on the market you know uber and google and amazon will fight for you right like you basically have to stand up raise your hand and the battle will ensue right so what what is the state of jobs in quantum for for these startups you know is there any kind of uh emerging resource uh because i've seen for instance academic consortia just plan to you know screed your boards and you know it's like it looks to me surprisingly uh underdeveloped in terms of jobs and matching uh candidates to companies absolutely and um i mentioned one quantum one of my three ventures which is a community organization and there we talk every day with uh dozens and dozens not every day with dozens and dozens but we have dozens and dozen startup members we talk to them every day um we actually just did a survey of what your top priorities are amongst those members uh surprisingly hr recruiting higher only came in a number four spot uh number one surprisingly to me but made me happy as a business kind of guy was a storytelling branding and marketing which i think 75 percent of people mention as a priority um then just you know sales and obviously fundraising um that said hr is a big problem everybody talks about talent shortage um almost every day i get an email you know saying i have a resume i have a job opening do you know somebody um and and it's both right it's companies looking for the unicorn post doc that that can transition into entrepreneurship and solve some of those problems that nick presented so so nicely but the business side is getting more prominent as well um more marketing needs to be done righty launched a new website yesterday which which looks fantastic um check it out if you haven't done so um but enterprise sales right just enterprise marketing which are things that we business folks have studied since the days of oracle 34 years ago how do you sell something to a jpmorgan how do you market something to a siemens all those kind of things um so those profiles um are being looked for more and obviously there are very few people that have that experience that quantum sufficiently um and and that's a big gap to fill as well so yeah if you're a postdoc and looking for something um um definitely let us know there's huge demand for that out there um if you're a business uh a person that wants to get into quantum there's a huge demand um out there for that hit up people like uh terrell um france who is in the audience um to figure out how to educate yourself um and become smart you really need to understand quantum to be able to market and sell it thank you thank you any other questions well if not let's thank andre again excellent excellent talk i really appreciate it uh i guess we all learned a lot from from this and also there is this fantastic data resource which uh we can produce um uh any appetite for lightning talks it's been i know it's been a long set today so does anybody want to present and if so we can kind of decide i can do it if you're not too tired and i really enjoy nick's talk so it really complements and answers some of my questions that i was going to raise thanks so much what would be the topic of your lightning talk uh vqe are we there yet so i'll talk about my experience uh in the cdl i mean during one of the weeks we had interesting calculations for vqe and that's where i start kind of thinking about exactly the problems that nick was pointing out at some point sounds good i mean let's hear it okay so i will share with you a link um that you can post later this is for from my slide sharing and i will go now in presentation mode for my google slides so as i mentioned i had a great time um about a month ago at creative destruction labs um and the topic that i will cover is what is v and how can we use it um i know that most people probably are already familiar with that um let me figure out how do i hide this part can you see my slides yeah okay no i wasn't showing them okay uh so what are the problems uh prevailing uh preventing its utilization today and then i'll go over case study basically using uh ibm queue through tequila and penny lane and then i'll just kind of draw my conclusions and answer any questions uh briefly for those that are really really new vqa is variational quantum egg and solver it can solve different problems the whole point is that the hamiltonian is considered to be some kind of cost function and uh the steps in the process is to prepare some quantum state and evaluate the expectation value of this cost function in usually we talk about this in hamiltonian but when we have to talk to people about like a traveling salesman problem and around about clicks and other things we start talking about cost functions because this is typical for people from economics and combinatorics now when people who would like to hear more about that you can go to the slideshare net and they have a few talks about this basically that is once you prepare your state you can start rotating different parameters in your state in this case particularly it's very simple it's just rotation around y and angle you're rotating there i would change the parameters and this your trail function would evaluate again the expectation value and in algorithm you just go through each of these gates and trying to figure out with the classical algorithm what is the best uh next step and then you keep changing it now um and after what you shortly see in my calculations at the cdl uh i look up at the ibm and i found that they had this nice kind of graph where they can show that the exact energy and the vq energy are matching and this is pretty much the same graph that nick was showing for lithium and you can see the configurations and the link that people who are interested then you can actually go study the ibm and i really love the way the ibm is preparing the tutorials these are very useful things for people to practically do things but there is one caveat i mean this graph was computed with the simulator if you want to actually do actual simulations and do other things then things would depend on the implementation now in this same tutorial they go through simulating noise and then now you can see that there's some deviation from what would be the exact value and what you get with the vqe so you can see that there's about three four percent difference uh and this is something to to be expected with the noisy qubits that we have nowadays uh and i'm reminding you about the slide from nick showing that when if you do the calculations you don't get this nice curve on actual qubits and then you have to implement some kind of other methods where you do calculation and extrapolations so the same thing that i mentioned that i was doing at the cdl and this is the github page where you can access actually our work and for this week this was week two of the uh uh quantum boot camp uh my team and of course there were other things it wasn't only these things but of all uh the work that we presented to the rest of the people in the boot camp uh got the winning concord project um there you can see actually classical solutions of the same question the lithium hydrogen are using hartree fog then confined configuration mixing and so on uh now the heart reform you can see that's actually off the actual four configuration mixing solutions but it's important to to always compare to see where it should be uh if you actually use the tools that we were playing with you can create a variation of circuits and this is just a snapshot of the variational circuit at one point uh and what are the wave functions for one particular run um and these again i'm kind of pricing ibm because these are really useful things when you go to the queue experience and being able to see okay what was the circuit that was executed and what was the outcome of the shots there now uh what started this kind of uh exploratory project there well i was shocked that when i submitted jobs to the ibm queue we actually had to do hydrogen because for the lithium i think the 16 qubit was taken by some commercial run so we couldn't run we waited for a few days but we we couldn't complete this runs and we decided okay let's go and do uh a hydrogen because that was the next one we could do and then i noticed the same thing that nick was pointing out that you actually get quite far from where you should be and there are different methods actually now uh the blue dot is where you start your algorithm to approach to the value that you would like to do and there are different approaches like unitary quantum coupled cluster method or coupled cluster method quantum quantum clipper method or you can just submit the job to the ibm um so it seems to be far away from where you want to land and again here some of the links what we actually learned was i mean this was used with the killer uh package and we found that definitely they had a problem with their private key type implementation so eventually we had to do the jordan wigner um and once we identified that there was an issue we thought well maybe tequila just screwed it up and we look at the penny lane and we did a little bit different we kind of randomly we knew what should be the value and then we submit to penny lane actually we submit first to ibm with tequila then we submit with the same parameters to penny lane and then we go back and submit again to ibm and you see that you are all over the map this is uh not controlled like what nick was doing we have a parameter to access we just wanted to confirm that uh definitely these functions within tequila and pennylane they were not set up already uh to take into account the noise and the calibration um so i was thinking well maybe if we look at the calibration and we can fix it but i understood from nick that calibration wouldn't help you much so in conclusion i would like to point out that well one has to cross check the results and then cross check again um and hopefully um i mean if we use the proper error mitigation and uh extrapolation eventually we will get good algorithms to uh utilize what is currently available as quantum devices so with this i'll stop here and get any questions from you again here's the github you can access uh that's the original cdl quantum and then if you want a little bit more victory or other talks that are given then you can access them through the slideshare um with this let me see if there are any questions so probably if you say then they'll hear it thank you wesley that was a great lighting talk and an example of what it should be you know very concise very clear to the point of the github link fantastic thank you yeah any questions uh hey i just like to say that was a that was a really great talk and like you really touched upon how a lot of decisions that you make in these applications you know really matter so i i think i just presented like a handful and you really kind of rounded out a lot of the rest of the choices that you need to make and so one of these things we've realized for researchers in the field is that a lot of people need to learn you know even even if you've like you kind of know how quantum computing works there's a lot of choices that are out there there's a lot of ways you can map your circuits to the qubits there's a lot of um you know choose your optimizer choose your uh choose your onsets choose all these kinds of things and and so there's a lot out there to consider yeah i agree it's and that's a lot of fun when if you're doing it right so i'll stop sharing my screen yeah thank you very much any other questions well if not i think we just concluded our third quantum conversations thank you so much guys for all the participation uh i think it was a smaller group but i feel like it was really more intimate and kind of more communal we really got through you know more interesting uh set of introductions and learn more about everybody so i really appreciate your participation the video will be as usual on function.tv and we'll continue this cadence we'll probably shift it a little bit earlier so it used to be the last wednesday of the month i probably will move it a bit up so it will be probably penultimate wednesday of the month uh so i'll send the invite as always on the list and quantum dot sv is the resource so i got the links finally from uh alex and wesley and i'll share them uh as well on the site thank you very much and uh until the next time see you thank you bye thanks alexi thanks everyone you