Venture with Grace
00:01:19 · Juha Vartiainen
[00:01:19] S1: Right. So maybe I start from year 2001, so twenty five years ago. I should say that that was the beginning of my journey in quantum computing. There was a Japanese professor, Mike Nakahara, who came to Finland as a visiting professor, and he gave lectures in quantum computing. That was first time ever in Finland. And I was his student on those lectures. That was exciting. I was a student that time. I graduated and I needed a topic for my PhD thesis. I actually selected quantum computing. So 2005, I made my thesis on quantum computing and Nakahara was my adviser. But he went back to Japan and kind of remotely advised me. Right. So then I left the academia. I thought at that point that quantum computing is not going to work. It's too immature. It was early stage of this technology. And for twelve years, I worked in another industry. I was in international standardization. It was the days of Nokia in Finland. And back in the days, there was like... Was working in a space industry, earth observation, satellites, data processing for those ones, R and D for some healthcare industry. Then after, let's say, 2015, there's had to be like venture capital flowing to Quantum Computers. So around 2017, 'eighteen, we started discuss the idea of founding a quantum company. And then I left my job and we become an entrepreneur together with my cofounders. Then we... Like 2018, we registered the company. In 2019, we got our seed funding, 11,500,000. And since then, we have been ramping up the company. And it has been a great journey.
00:03:31 · Grace Gong
[00:03:31] S0: For sure. I feel like there is technically like two giants in the quantum space. One is you guys, one is kind of like IBM. And I want to start with like, you know, for people who first time hearing about quantum, like, what exactly is this? So from like a normie perspective, from my explanation of quantum is essentially in the past, computer can only process one set of data at once. But quantum computing kind of help you to process a bunch of stuff at the same time. But I want to hear your explanation on what is quantum and then how is it differentiated from traditional computing?
00:04:07 · Juha Vartiainen
[00:04:07] S1: Yeah, there are actually many ways to explain it. One maybe important piece of information is that quantum computer is not like a faster laptop, not at all. So it's a very special kind of computer. It's more like an accelerator, like GPUs are. So it's particularly good for some type of computations, not for general computations. And as you said, it uses different laws of nature. So that's why it can do magic. So it uses these laws of quantum mechanics where, let's say, atoms can be excited and not excited state at the same time. So similarly, we can put the electric circuits in superconducting chips to be in a state zero, state one and different combination of those at the same time. And actually, we can in one step solve a problem which would take a regular computer to go through all the combinations. And a little bit like this, following this idea, it can potentially in the far future break encryption code. So it can solve some particular mathematical problems and it can change the complexity class of computation. So it scales in a very difficult way in a regular computer, quantum computer in some cases can solve that problem very easily. And there are some particular type of use cases, but not internal, not everything. So usually we see nowadays like the quantum computers are there in a data center on side of high performance computing, like AI factories, where there are some racks of CPUs and GPUs, and then we hope there will be rows of QPUs in the same data center for particular purposes.
00:06:15 · Grace Gong, Juha Vartiainen
[00:06:15] S0: Speaking of QPUs, I want to talk about, there is a term that I feel like you're very good at explaining based off how I was listening to your past interviews why it's called superconducting qubits. Do you want to explain this to the general public? [00:06:32] S1: Okay, superconducting qubit, it's basically, it's a resonator. So it's like a tuning fork. The electrons go back and forth in a qubit 5,000,000,000 times a second. And superconductor by nature doesn't have electric resistance, so it shouldn't die at all. But since there are some small imperfections when they have gone like 5,000,000 times there, So let's say half of the resonance have died. So let's say it doesn't last forever, but it can last like one millisecond. And in that time, we have to do all the computations. So that's basically the thing. So if that would be a tuning fork, I would hit it today. It would be still ringing next week. So it's like a very good resonator, but it's not perfect. And what we are going to do is we try to make it perfect. So we want to do the material science and do our best to remove all the imperfections and make it live forever. And once we manage to do that, then we can code more complex computations into these qubits. So we can set them to be in the state zero, state one, and the superposition of those ones, preferably forever.
00:06:32 · Juha Vartiainen, Grace Gong
[00:06:32] S1: Okay, superconducting qubit, it's basically, it's a resonator. So it's like a tuning fork. The electrons go back and forth in a qubit 5,000,000,000 times a second. And superconductor by nature doesn't have electric resistance, so it shouldn't die at all. But since there are some small imperfections when they have gone like 5,000,000 times there, So let's say half of the resonance have died. So let's say it doesn't last forever, but it can last like one millisecond. And in that time, we have to do all the computations. So that's basically the thing. So if that would be a tuning fork, I would hit it today. It would be still ringing next week. So it's like a very good resonator, but it's not perfect. And what we are going to do is we try to make it perfect. So we want to do the material science and do our best to remove all the imperfections and make it live forever. And once we manage to do that, then we can code more complex computations into these qubits. So we can set them to be in the state zero, state one, and the superposition of those ones, preferably forever. [00:07:57] S0: So nowadays, I feel like since you briefly mentioned about the data center and now maybe there was quantum data center. So basically, you frequently talk about quantum AI and HPC. I want to talk about where... So how is AI helping building better quantum computers? So obviously you guys recently announced like an AI driven agentic collaboration with NVIDIA. So versus quantum helping AI? How does AI shape quantum computing and how does quantum kind of help AI?
00:08:38 · Juha Vartiainen
[00:08:38] S1: Right. Are are many ways engineers So are using AI nowadays, many ways, but this may be the most interesting ways to use it for calibration. So actually quantum computer has like a lot of knobs to tune. So And the performance of quantum computer requires that it's properly calibrated. And there are many ways how it can go out of like a perfect operation point. And traditionally, it used to be like PhDs adjusting some knobs and putting it back to the perfect sweet spot. But now when the quantum computers get more complex, there are too many parameters to calibrate for a human. So there are automatic calibration routines, which are computer programs which are having certain rules and they take certain steps. But then when the complexity of quantum computer grows, there are always some failure modes. It's very difficult to anticipate all the possible ways what can go wrong. So sometimes even in a perfect algorithm, it fails and we need to call a human to assist. But the problem is that it's not scalable way. So we can, of course, have one quantum computer. We can have some dedicated experts night and day keeping it working, but that's not scalable when we are selling more and more quantum computers that are sort of have an equal number of PhDs with education there to keep them calibrated. So fortunately, we have now the AI. So actually, NVIDIA has developed a very nice set of open source models. And particularly for this purpose, there is NVIDIA icing model where this calibration data has been fed to the visual language model. And actually, the AI model can watch the data. So we show the AI the data. This is what we measure from the quantum computer. Is it okay? It's okay. Just keep on going. Or it can say that, okay, actually, it seems that not everything okay. Please, let's adjust this and that parameter. And then we do the adjustments according to the AI's instructions. And actually, it's surprising that that way we managed to keep the quantum computer calibrated. So our latest result we were discussing in some scientific workshop lately is that our 20 qubit quantum computer was fully calibrated using the agentic models.