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 updates from the quantum computing hardware landscape. For this, I'm very honored to have a special expert guest and fellow Microsoft MVP and Microsoft Regional Director, Ciprian Jichici. Welcome to the show, how are you today?
Ciprian - Hi, I'm good. I'm very excited to come back to your show. Thank you very much for having me. And, of course, the topic is a very interesting one because every month we see some new announcements, and things are really getting hot in this space.
Rene - Before we start, for folks that might not have seen you in a previous episode, can you tell us a little bit about yourself and your background, and what you're doing with quantum?
Ciprian - Absolutely. One of my primary areas of work is leading the advanced computing practice at Salinas Alliance, which is one of the Microsoft Global partners. I do a lot of quantum computing-related work over there, mostly in the space of advisory, but also lately we have been involved in projects related to post-quantum cryptography, the security impact of new quantum computers, as well as a couple of interesting projects where we explore using quantum computing to solve optimization problems. I like to call myself a failed physicist in a sense that 25 years ago when I started my career, I had to choose between physics and computer science. I chose computer science, but interestingly enough, fast forward 25 years and I find myself again working in both areas. So, this is really exciting. And, as you mentioned, I'm a Microsoft Most Valuable Professional for AI and for Quantum Computing, which is very exciting because it allows us to work directly with the Microsoft Quantum team.
Rene - And don't forget to tell our audience that you're also giving a course at a university.
Ciprian - Yeah, this is something new and it's also very exciting. This fall, I started an Introduction to Quantum Computing course at my local university in Timisoara, Romania. I can tell you I was really surprised about the level of interest we see among the students. It's not only that the interest exists in the industry, but it seems the interest exists in the education space, which is very encouraging for me, I have to say.
Rene - Yeah, it's really impressive. In fact, for Season 6, Episode 1, which we're recording right now, I interviewed a student from the US who started with quantum when she was 15 years old. It's just super exciting. It's really something that the young folks are very eager to learn.
Ciprian - And, it's very interesting that you say that because I think two or three months ago, on the podcast I do with Patrick Heinz called "Entangled Fact Things," we had a guest who was teaching an Introduction to Quantum Computing in one of the high schools in New York. So, it's not only seen in universities but starting, obviously, the level is dialed down to allow kids from high school to understand the concepts. But this is actually what I call the quantum first generation, all of us including me and you, who have a bias for classical computing because that's where we have grown up. But we see more and more of this generation who has its first contact ever with computing in the form of quantum computing, and that will be an interesting generation to watch in the coming decade.
Rene - Absolutely. And, yeah, like you mentioned, you also run a podcast with Patrick Heinz called "Entangled Things”. Awesome! So, check this out, folks. We have some good podcasts here. I mean, there are a few more, but the best ones are Entangled Things and QuBites, of course, so check those out. All right, well, let's get started with some of the questions. And you know, in season two, episode two, we talked about the different kinds of quantum computers that are out there and the different approaches for building the qubits. I keep on saying 'QuBites,' but it's for realizing the qubits. And you know, which ones were the most promising in your opinion. This was in February 2021, which is quite a while, one and a half years ago. So, yeah, what has changed since then? Any good stuff there?"
Ciprian - Well, from my point of view, I believe we see a consolidation in the top two modalities that yield the highest quality practical results today. I'm talking about trapped-ion and the superconducting approach. We've seen a few releases of new chips in this space, and an interesting development with respect to Microsoft's topological quantum computing approach. After a major setback in 2018-2019, it seems they have recovered significantly. IBM recently launched a 433-qubit processor, which are physical qubits. IBM has a timeline where they expect to reach close to 1000 qubits in 2023 and potentially up to 2000-3000 qubits in 2024-2025. We are starting to see clarity in these modalities, with trapped-ion and superconducting having practical results. Microsoft has also proved that their topological approach has validation in theory and practice. Alternative modalities, such as nitrogen vacancies in diamonds, provide interesting types of qubits, but they are difficult to manufacture at this level. The challenge of scaling up is confirmed for both superconducting qubits and ions. Moving from a small number of logical qubits to a large number proves to be a significant engineering challenge.
Rene - Yeah, absolutely, and like you're saying, IBM is really pushing forward with the highest number of qubits, but they're all noisy or physical qubits. How much do you think is actually required to make a logical qubit that is more stable? What is practical and usable if you have a 1000-qubit computer?
Ciprian - At this point, we still see some significant error rates ranging from 10 to the power of -3 to 10 to the power of -2. This means that thousands or hundreds of physical qubits are needed to get logical qubits. The interesting thing is, I had another guest, a professor focused on error correction in circuit-based Quantum Computing, who told us that 2022 and 2023 will mark the tipping point of quantum error correction becoming a positive factor in Quantum Computing. Up to this point, any error correction required more noisy qubits and the noise of these physical qubits actually decreased the quality of the computation. However, it seems that the stability of qubits has reached or will soon reach a point where error correction will actually start correcting errors and not decrease the stability of the system. Although we're still far from super stable logical qubits based on error correction, this is a remarkable point of inflection in the evolution of both the hardware and the error correction codes. There are also a couple of interesting proposals for error correction in quantum computing without codes, which is an interesting approach. The classical computing-inspired approach uses codes, but there are alternatives. I think there are a lot of things moving in this space and I believe we're on the path of seeing a better ratio of logical qubits to physical qubits required to produce them.
Rene - Got it, yeah that's actually a really exciting development. We also mentioned that there are different approaches where you can have more stable qubits, and we already mentioned Microsoft's approach with the topological qubits which is meant to be more stable and leverages something. You also mentioned they had this breakthrough with what is called the Majorana zero modes. Can you explain what it is and why it's such an important milestone in fact?
Ciprian - Sure. Microsoft's approach is a very interesting one in the sense that it is unique. It's neither based on the classical superconducting approach nor the trapped ion approach. Essentially, what they are building is, if we oversimplify the whole thing, topological superconducting wires. These are very, very narrow areas of material that have superconducting capabilities but also exist in a very specific state of matter, which is called the topological state of matter. Without getting into the details, this is essentially a state of matter where matter behaves more like in a 2D style rather than a 3D style, and it's more stable. It's easier to keep the stability. So imagine these wires, and at the end of these wires, they have these energy gaps, which can act as qubits. These are the so-called Majorana zero modes (MZMs). What Microsoft was actually able to prove earlier this year was that the gap between these spikes, these Majorana zero modes, was around 30 micro-electron volts. This number in itself may not be interesting, but what's interesting is that it's actually three times larger than the average noise in the experimental environment. This means, in plain English, these Majorana zero modes that they found seem to be very well isolated from the background noise of the system, which is one of the fundamental problems that all Quantum Computing building modalities have to deal with. This is why the result is remarkable, but the other thing I would like to point out is after the setback Microsoft had circa 2018-2019, what they did this time was truly remarkable. They actually published all their experimental results. There is, in fact, a notebook that anyone can download that contains a huge dataset with all the experimental results. The results are peer-reviewed, and anyone can download the results, do the scrutiny on them, and potentially even review some of the claims that Microsoft made. And I find this truly remarkable. And also, it's a very important step because it seems to be a major shift in Microsoft's strategy. According to what they say, they plan to continue releasing their experimental results. So, it's encouraging. If you think about superconducting and trapped ions, those were the low-hanging fruits, while the topological superconducting wires are a much higher-risk approach. However, if Microsoft manages to build a stable qubit based on these, their path to scaling up, from 100 qubits to thousands, will likely be smoother. This is exciting and it's being watched closely.
Rene - The fact that the results are open source is even more exciting, as it allows the whole community to build on top of it, just like with AI in the last couple of years.
Ciprian - Microsoft is a platform company, so while they are driving their hardware quest, they are also building a full stack that ranges from the capability of incorporating third-party hardware, to building simulators, programming languages, and full environments for quantum computing. So now, you can choose the Microsoft way or the IBM way, but what's important is that if you want to start learning quantum computing, you already have environments where you can start testing your ideas and algorithms.
Rene - Yeah, you can actually run a bunch of simulations as well, right. In fact, I was also talking with Jin, the developer relations manager at Nvidia, and he mentioned a couple of things. Interestingly, I think I saw it at Nvidia GDC before, they have these huge DJX A100 systems, like crazy launched GPU systems, and they were able to do what is called a superpot. You need a lot of resources, of course, a lot of GPUs, and they can simulate 5000 stable qubits, although it's just a simulation. This is something similar that Microsoft is also offering now. And yeah, you can already get started with quite a lot of qubits.
Ciprian - It's important to mention some important advancements that D-Wave is making. Compared to the other companies mentioned, D-Wave is building quantum computers for adiabatic quantum computing, which is a different approach. The latest one they announced is 4000 qubits for adiabatic quantum computing and it's increasing. These are more specialized types of hardware that are geared towards solving optimization problems, but the level of scale is impressive nonetheless. I have to say, D-Wave is also doing some remarkable work and they're moving fast towards their first big milestone, the 10,000 qubit chip for adiabatic quantum computing. However, they're facing similar problems in terms of cooling down the chips and scaling up. One of the big problems of scaling is providing the connections between the qubits. It's easy to apply single-qubit gates to the qubits, but problems start when you need to apply two-qubit gates because then you're starting to depend on the topology of the connections between the qubits. Not all pairs of qubits can actually be subjected to two-qubit gates, which is a significant limitation in terms of implementing algorithms in a practical way. This is a fascinating aspect of building the hardware. There are many things you need to address, not only cooling down and reaching insane levels of finesse in terms of engineering, but also the topology of the connections, which is a fascinating problem. These folks need to address this.
Rene - Absolutely, and I was recently talking with a quantum chemist who works in chemistry and they are using D-Wave systems for drug discovery. They are doing the initial phases of protein analysis and it works quite well for them because it is much faster. They can also build things like personalized medicine, which is a visionary and impactful topic. She told me that with their approach, they can do the initial phase of drug discovery in six months or less, which used to take a few years. This is impressive and could save someone's life. You mentioned about players that are researching about Quantum Computing. Any player that we missed?
Ciprian – The list is fairly long. Obviously we will have to mention Righetti is another interesting player and so is the spin-off from Honeywell, called Quantino. IronQ is another interesting player, as well as Toshiba and Intel. The field is attracting more high profile names because now the industry is starting to see practical results. With optimization processes being run on adiabatic quantum computing and circuit-based quantum computing processors becoming larger than the five qubits of a few years ago, interest is starting to increase. There are also a lot of combinations and collaborations between private companies and academic spaces, with major universities around the world running quantum computing programs or entire departments. It's refreshing to see that we have options and are still on the non-hype side, although we still have a long way to go. We're not yet at the point of breaking RSA 2048, just to be clear. A lot of investment still needs to be made, I would dare to say a lot of innovation still needs to happen, but the field is moving, and it's moving at a very fast pace. At the end of the day, no matter how many smart algorithms we design, if we don't have the chips, the physical devices to run them on, they will be useless. One other interesting topic or modality that we haven't talked about is the creation of qubits systems that will run at higher temperatures. To keep things in perspective, current qubits run at tens of millikelvins, which is 100 times cooler than the coldest place in the universe. These new approaches are trying to implement what we call "room temperature" qubits, but they will probably run at five to ten Kelvin. This is a very interesting line of work because to reach the temperatures of tens of Millikelvins, the infrastructure required is extremely complex, and the coherence times of those qubits are very short-lived. The propositions of increasing the temperature at which those qubits live are very important. The higher the temperature, the more movement you get in the system, and the more noise you get. Ideally, we might have qubits that work at room temperature someday, but we don't even know if that will ever be possible.
Rene - That was fantastic, Ciprian. Thank you so much for joining us today and sharing your insights. It was fantastic to hear all about it. Very much appreciated.
Ciprian - Thank you very much for having me on the show. It was a real pleasure.
Rene - Well, thanks to everyone for joining us for yet another episode of QuBites. Your bite-sized pieces of quantum computing. Watch our blog, follow our social media channels to hear all about the next episode and of course, subscribe to our YouTube channel to get a notification when the next episode is available. You can also find all previous episodes, including Season 2, Episode 2 with Ciprian, and all the other guests, on the website. Thank you so much. Take care and see you soon. Bye bye."
