Transcript
Rene – Hi, welcome to QuBites, your bite-sized pieces of quantum computing. My name is Rene from Valorem Reply and today we're going to talk about the intersection of quantum and classical computing. And for this I'm honored to have a very special expert guest today, Robin Coxe. Hi Robin and welcome to the show. How are you today?
Robin - Hi Rene. I'm doing great, thank you! Thank you for having me.
Rene - Awesome to have you. So let's start with the typical introduction. Tell us a little bit about yourself and your background as relates to quantum computing, engineering, whatever.
Robin – Sure. My name is Robin Coxe. I am the Vice President of control systems engineering at Atom Computing and I originally have a background in physics but not in quantum information. It was in experimental particle physics I have a PhD from the university of Chicago and I also studied physics as an undergrad at Harvard.
Rene – Wow! Well, you already hinted at this right. You're working on control system, basically on the intersection between the quantum computing and the classical computing kind of. But let's talk first of all about some of the stuff you do at Atom Computing and for making sure that the whole audience, also me, understands what you are doing and what's so special about your approach there. And in season four episode seven, I actually talked with your colleague Denise Ruffner and she told us about her work at Atom computing of course. But for those that have not seen the episode, can you tell us a little bit about Atom computing and what are the benefits in this so-called neutral atom approach that you're using at Atom computing?
Robin – Sure. So Atom computing was founded in 2018 by Ben Bloom and Jonathan King and we now have several locations but our original location is in Berkeley, California and Ben is an expert in atomic clocks and Jonathan in nuclear and magnetic resonance. And we're building quantum computers in a slightly different way, where they're based on nuclear spin qubits and we make these qubits by cooling and trapping alkaline earth atoms one by one in a two-dimensional array and we manipulate individual atoms with a suite of lasers. So, lasers as we'll find out, as I'll talk about a little more in the future and you know coming up these lasers are controlled by what way by radios and that's where I come in, because I have a lot of experience, after I left grad school, I worked in telecom and wireless for a long time. So a lot of that technology has a direct relevance to what we're doing. We're growing quite rapidly. When I joined two and a half years ago, I was employee number nine. At this time last year we were 20 people and now February 2022, we're close to 50 people. And last summer actually we announced that Phoenix, which is our flagship quantum computer, surpassed the 100cubit milestones, so we've managed to trap an array of more than 100 atoms. But we can make qubits out of 100 of them and last month actually we just closed our series B round eventual venture capital. So that's very exciting and we're planning on building two bigger and better quantum computers in our Colorado office which opened last summer. And are preparing to open up those machines to external users. So, that's where people like Denise come in. She's our Chief Business Officer and she's working hard to put together some partnerships and customer engagements for when we open up from our machine to users.
Rene – Awesome! So is it some kind of a quantum as a service offering you're looking at? Basically that people can rent computing time on your machine?
Robin - That is certainly one model and we're kind of brainstorming other ways in which we can provide people access to our machine but I did want to talk a little bit more about why neutral atoms. Because you know, when people see quantum computing in the popular press, they might have seen like IBM or Google has the big thing that looks like refrigerator and cryogenic temperature, so those are called transparent superconducting transmon quantum computers and so it's a completely different approach we're using. Atoms, so neutral atoms are by definition identical to each other. So we don't have to contend with the process variations that people making super conducting qubits have to deal with. It's very hard to make identical structures on a, you know, a semi-conductor plate or what have you. And so, we don't have to do that. We just, you know, we just had all strontium atoms or all your terbium atoms are identical. So, we just need more atoms. Also, nuclear spin qubits, because we're kind of using the magic of atomic physics to turn them into cubits. They have much longer coherence times and what I mean by coherence time is essentially the period in which the qubits can retain useful information. So, neutral atom qubits can persist for like tens of seconds, which is orders of magnitude longer than some of the other technologies like superconducting transmon or trapped ion qubits. The flip side of having long coherence time is the computations take place at a slightly slower time scale. So you know it's like having a CPU with a slower clock rate but the each state persists for longer. And you know, another huge advantage of neutral atoms is the number of qubits that we can in principle attain in our computer scales approximately with the laser power. So, the scalability of neutral atoms is quite promising compared to some of the other technologies and the qubits themselves are smaller spatially. Smaller than some of the super-conductor based qubits and you know since we use the lasers themselves to actually cool and trap the atoms, having a big cryogenic system is not an absolute requirement. It might, you know, you can have a cryogenic based neutral atom quantum computer and there's some advantages for doing so but because we don't have to have a lot of electronics in the cryogenic environment, it makes the system a lot kind of simpler. Another advantage from the electronics point of view is we can address individual atoms but we don't need a single channel of electronics for each atom unlike, so, for instance, for a superconducting transmon qubit you need one individual electronic channel to excite it and another one to read it out whereas we can address multiple atoms in the array with many fewer channels of electronics. So the amount of stuff we need is a lot smaller as well and it makes the system a lot simpler and we do the readout actually we can read out the entire array by taking a series of pictures. So we're essentially, instead of reading out using rf, like some of the other techniques used, we take pictures with a high resolution scientific imager. So our system architecture is quite different in some respects. But the scalability, the promise of subsequent machines with many more qubits faster is one of the more appealing aspects of it and I just like to say, when I started at Atom computing, I knew absolutely nothing about any of this and they explained it to me and I do have a physics background but I did in a different field of physics but it actually sounded plausible to me so I didn't like turn and run.
Rene - And you also explained it super well to me right, now work to our audience. I get it now and it's also pretty amazing like you're saying when you do the readout you're basically leveraging image processing to read it out that's super clever. Like I said like this approach scales well because you need lasers, basically, to cool down and you know I don't need these huge cryogenic chambers and all of that stuff
Robin - And we are benefiting from some extremely clever atomic physics techniques using lasers that were developed over the past 50 years or so in academia. And we're the directive beneficiaries of all of this amazing physics too. So it's been fun for me as you know someone who's studied physics, I haven't really been a working physicist for a long time but now I'm surrounded by them. So it's very exciting work environment.
Rene - Love it. Let's talk a little bit more about the tech thing. I think the challenge is basically when you work on the electronics on the control systems. Why is it such a critical component of a quantum computer and what are the challenges really to build this as a robust controller unit? Basically because you still need classical computers, right? You need controllers and all of that to actually control the quantum computer which a lot of folks still think like we don't need classical computers in the future anymore. Well that's not going to happen of course right? It's basically you know always keep on saying basically the quantum computer should be more seen as an accelerator chip just like a GPU or whatever.
Robin - Certain types of computational problems are much more amenable to quantum computers but yeah classical computers are not going to go away anytime soon because we actually need them to make the quantum computer work in the first place. So that's a good segue into kind of what my group does. So the control systems team at Atom computing, we call ourselves the non-quantum engineers and we inhabit kind of the bottom part of this computing stack that's closest to the atom. So, quantum circuits can actually be represented as sequences of radio frequency or rf pulses, so more specifically you know, I mentioned we rely on a sophisticated suite of lasers to like cool and trap and manipulate our arrays of atoms into qubits and in order to actually do that we are manipulating the atomic states of the atom. So we need to be able to control the amplitude, the frequency, the phase and the deflection angle of the laser light very precisely because we and the way we do this is we input our radio frequency pulses into these devices called electro-optic modulators acousto optic modulators and acousto optic deflectors which interact with the laser light and manipulate it in the ways that we want it to. So we can cool the atoms and then put them in their 2d array and then make our qubits. So what the fundamental part of our control system element of our control system is something called a software defined radio and this thing is a radio with multiple transmit channels and the software defined part all that is referring to is like the ability to generate specialized wave forms digitally in software so we can define them using basically computer code and then they're converted into radio waves in the rf electronics chain that we also build. So, fortunately we can use many of the commercially available integrated circuits that were originally developed for like wireless and telecom networks and links between you know satellites and earth because that's all over radiofrequency links as well and so we can build a software-defined radio with like eight outputs on like a single card that's like the size of a clipboard basically. And the way we scale up the number of channels that we can support is just like putting more cards into into a chassis. So the analog is like in a big server farm of ig servers. If you want more capacity, you just slot in another server into the rack. It's a similar idea here and so our system right now, our phoenix system, was originally built with a whole bunch of commercial boxes that were like piled on top of each other but if we want to scale that up you know, from hundreds of thousands, we need this kind of line card architecture. So my group is developing some kind of semi-custom electronics to make this a reality and fortunately this is another kind of fun blast from the past, is people just designing control systems for particle accelerators use. You know they need to do some similar things so they've developed some commercial technology that we can leverage. So we don't have to design everything ourselves. We can buy a lot of commercial technology to give us a start on this kind of scalable line card architecture and so all this radio frequency where you know I worked in radio before and I'm actually using some of the chips that I helped design when I worked in analog devices. So that's fun. But the difference between our radios and like base station radios is we don't actually have anything wireless. It's all the all the radio frequency pulses are transmitted to these modulators and deflectors over like a huge collection of cables. So we need a lot of very expensive cables. So we don't actually have loss over the long cable but we don't have to deal with antennas or worry about interfering with the wi-fi or anything and kind of the fun. Another important thing that my group does is there we have these chips called field programmable gate arrays or fpas and fpgas have actually been in the news because AMD just bought Xilinx which is one of the major fpga companies and that merger just went through on monday and we actually use Xilinx fpgas in our systems and they enable us to hardware accelerate some of the more computationally intensive operations that we need to generate the exact waveforms that the physicists tell us are necessary to make the atoms bend to our wishes so to speak. And these fpga chips also have embedded microprocessors in them which kind of act as orchestrators in the data chain. So we have a bunch of different sort of digital signal processing steps that need to happen and we have like a little mini embedded CPU, kind of making sure everything is ..like, the signals are marching down the rf chain and the order that we need them to. So my team does a lot of kind of digital logic design and embedded linux and firmware development as well. So you know conventional classical ees definitely have a place at atom computing. In addition to the radio itself some of these RF output channels they need to be very precisely controlled like in a closed-loop way like in a thermo like kind of like a thermostat. Like if it's too hot we need to turn it down a little bit you know. If it's too cold we need to turn it up a little bit. So in other words we can adjust it like the amplitude of the frequency of one of the output pulses based on the signal response of the laser. So you know we put RF into the system and then we may we measure how the laser responds and we feed back the laser response to our software-defined radio devices and then adapt. So that's a big part of what we do as well. In addition to surveying radio frequency we also like ramp magnetic fields as well because we actually need magnetic fields to split the atomic energy levels in ways in the states that we need and again and we touch on also the readout. So we do some kind of very sophisticated real-time image processing as well on the readout side. The radio is mostly on the excitation side to get the qubits to be qubits and to do content computations and then the image processing is on right outside. So we do that's kind of the types of problems that my group is looking to solve and create into like scalable solutions that we can replicate for larger and larger systems.
Rene - That's quite a lot in fact. And very important pieces like otherwise the quantum computer in the middle wouldn't work right? So without it you cannot have it but let's talk a little bit more about the people. You already hinted at it that it's you not just need to have a PhD in quantum physics or electrical engineering and things like that and there's definitely this kind of workforce shortage in the quantum field and it has been also in our QuBites show, here a reoccurring topic and discussion topic if you will and I'm sure you're very aware of it as well and but in your opinion what are the kind of required skills actually to be working in quantum computing? Is it really a degree in physics, electrical engineering, computer science? What is the kind of stuff, skills required?
Robin – Sure. So although I happen to have one, I'd like to dispel the misconception that you need a PhD in physics. I didn't really use my PhD in physics until I got here and it was kind of useful to ramp up on the physics a little faster but it actually isn't a requirement for my job at all. There's a huge demand in quantum computing for skilled software firmware fpga and hardware engineers but it's not just technology. We also need to convince people that quantum computing is worth paying money for. So we need marketing and business development people who can actually understand the subtleties of the tech of what they're promoting or selling. So they actually portray it accurately and so customer expectations are in line with what the machine actually does. So that's becoming crucially important for us in particular as we kind of transition out of the science experiment phase and into the for-profit company phase. So that you know we've been a lot of our hiring recently has been on the product and the business side people like Denise and our chief product officer Justin Ging who's joined are kind of leading the charge in that regard and one thing that's there are also people in companies. So I think one thing that people don't talk about a lot which is like just crucially important is competent technical managers. So you know managing knowledge workers like the people say is and it's true it's kind of like hurting cats they're a lot of incredibly capable technical people but tech incredibly technical people are also humans. So I think you know this industry really could benefit from people with significant experience working at companies that are revenue generating and have made products, so they can stay focused and because there is a lot of hype in quantum computing. But we kind of have to really focus and we need to.. it's also kind of important to be first in a lot of respects. So we need to have people in companies who really understand the notion of opportunity costs. It's like you're working on task at the expense of working on tasks b, c and d and is that the best use of your time? So you know one of the things I've had to learn, you know, kind of transitioning from academia is perfect, is kind of the worst enemy of good enough like it's good enough and it meets the requirement. You know actually optimizing it more is not a good. We've had this kind of internal discussion in our company that incremental improvements even to the number of qubits are not going to cut it in the long term. We kind of have a limited time to ride the next big thing with technology and I think we need to swing for the fences and I realize I'm using a sports analogy for a sport I don't even like that much which is baseball but I'm a soccer fan as you can see the soccer ball right there or football as the rest of the world calls it. Anyway as far as qualities that will you know that will make someone successful in a quantum computing company or like any technology company is you know being very good at your technical field. If you're you know a stem person attention to detail, thinking systematically, the ability to teach yourself stuff because we're doing a lot of stuff that you might not have learned in school, particularly if you're not a physicist and just the ability to sort of seek out experts, read up on things, ask questions, and I mean one of the things that I think we've done a really good job of at Atom computing is hiring incredibly smart people who are also humble. So not letting your own ego and get in the way of sort of collective progress is I think really important. And one of the things that I think is generally not stressed enough in technical fields is the ability to read and write clearly so whenever like people give me like physics undergrads ask me for career advice I'm like, being able to get give good presentations and you know write it you know to describe what you're doing very precisely like you want non-experts to understand what you're doing and you want your bosses to be able to kind of recognize your contribution and your values to your organization and like they're not necessarily going to be experts in what you do right? So I think that's super important and I think as managers in these organizations I think I've definitely found that people crave transparency and just they want to know what's going on? They want to know why we're doing certain things and making sure that why everyone understands kind of the end goals of and as a manager and like saying and making it clear like who is responsible for what and who owns what I think sometimes if someone says we should just do this and no one knows that they're supposed to do it. So no one does it. So I think particularly in rapidly growing companies where you're kind of like ready fire, aim right? We need to sometimes just hit the pause button take a step back and be like why are we doing this? You're responsible for it and we'll check in a week to see how it's going and if you meet in a roadblock which happens in technology and doing things that are new then we change course a little bit but it's not the end of the world right? So I think a lot of these things that i'm talking about will not just make you successful in quantum computing but just like in general in a job so I don't think it's specific to quantum computing but I think like there's nothing that, oh but on the flip side there's nothing that highly skilled technical people hate more than micromanage, that being micromanaged. So I think just managing with a light touch and caring about people as human beings goes really far. I think if quantum computing companies specifically can establish reputations amongst employees that they're not dysfunctional or toxic places to work recruiting will be easy because you know everyone likes the idea of working on the next revolution in computing, like it's very exciting when I was looking for a job before I joined Atom like the other offers I had were kind of with convent more conventional wireless companies like what's the next opportunity to like it might not workout but like it'll be interesting and the skills that people on the control systems team have or pretty much everyone who works here has are very transferable not only to other quantum computing companies but also to other industries. So I don't you know some people view joining quantum computer companies is risky but I actually think it's not at all. I think it's like a great opportunity and like even if your individual company doesn't succeed you know I think Adam computing is going to succeed or I wouldn't work here but even if it doesn't I'm not worried. And I'm not worried for any of the people who work here honestly.
Rene - So there's a diverse skill set needed and also what you were saying is you need to be also a good translator right you need to be able to translate these well sometimes hardcore technical terms into things that you know people can understand. Especially if you want to widen the scope of the folks you want to work with right you need to be able to basically explain it by looking at your audience and thinking about your audience and thinking about their background and this is the way how you should approach it. if you if you talk about a quantum and you know like I love the advice you gave because like the same advice I'm also giving often to students and so on that they are like really not just need to focus on the hard skills if you will but also on the soft skills part and especially being a good translator right? Reading but also speaking and explaining these things in such a way that it can when I say widens scope I mean like reach more people and rate them.
Robin - My colleague Denise runs this organization, woman in quantum and I really applaud her initiative in starting that and I definitely am committed to kind of doing my part to encourage typically under represented people to join this industry. I think you know, I definitely think that if you can see it you can be it effect is very powerful like, people see people like themselves and maybe think oh maybe it won't be exclusionary or I'll fit in and that kind of thing so I and I think it's sometimes hard to quantify the value of all forms of diversity but like I think sometimes people kind of naturally surround themselves with people who are like them and lose sight of the fact that diversity of thought actually will naturally kind of bubble up from having people of different backgrounds and perspectives. Like in the room and given a seat at the table so I mean like we have discussions all the time like that we don't want our company to just be like tech grows , business grows because you know you fall into like toxic group thing and then you kind of put your blinders on and you realize how much you're missing and how much you're not seeing and I feel very strongly that like that's not good overall like financial viability of an organization either because you're just missing out on opportunities to take advantage of people's skills and talents and also mark you know market your product to a wider swath of customers too, right, because I think a lot of people make like purchasing decisions about like is this product like consistent with me and if it's not then like you're not going to pay money for it.
Rene – Right, awesome stuff, Robin. We talked for quite a while but unfortunately we're already at the end of the show. We could continue for much longer. You're such a great guest here and explain all these things so well and with this lens also, with your other experience, right, coming from a classical electrical engineering background basically. It's super appreciated that you have been with us here today. Thank you so much Robin.
Robin - Thank you!
Rene - Well and thanks everyone for joining us for yet another episode of QuBites, your bite-size pieces of quantum computing. Watch our blog and follow our social media channels to hear all about the next episodes and of course you can watch all previous episodes from season one to season five, also with Robin's colleague before on our website you can find all these episodes. Again thanks so much, take care and see you soon, bye!
