Tommaso Demarie (PhD) is co-founder and CEO of Entropica Labs, a company working on optimisation with and for quantum computing.
Tommaso has 12 years of experience in the field of quantum information and computation. Before founding Entropica, he was a Postdoctoral Research Fellow at Singapore's Centre for Quantum Technologies and Singapore University of Technology and Design.
His research interests included (and include, when he finds the time to publish research articles) applications of quantum-based protocols, verification of quantum computation, and novel cryptographic schemes for quantum computing.
Tommaso holds an Undergraduate Degree in Physics and a Masters in Environmental Physics from the University of Torino (Turin, Italy) and a PhD in Physics from Macquarie University, Sydney.
Tommaso's Google Scholar
Tommaso's LinkedIn profile
Michael Liebreich: Before we start, if you're enjoying these conversations, please make sure that you like or subscribe to Cleaning Up, it really helps other people to find us. Cleaning Up is brought to you by the Liebreich Foundation and the Gilardini Foundation. Hello, I'm Michael Liebreich and this is Cleaning Up. By now we've all heard of quantum computing. But what is it and what will be its impact on the net zero transition? My guest today on Cleaning Up is Tomasso Demarie. He's the CEO and Co-Founder of Entropica Labs, a company working on the use of quantum computing to solve real world problems. Please welcome Tom to Cleaning Up. So, Tom, very good to see you here.
Tomasso Demarie: It's great to be here. Thank you, Michael.
ML: Now, tell me where are you at the moment? Where are you dialing in from?
TD: So right now I am at home in Singapore, which has been my home for the last eight years. In fact, I came here to work at the Centre for Quantum Technologies at the National University, and I'm calling you from there.
ML: And is Singapore a hotbed of quantum computing?
TD: Great question. Singapore was very actually forward looking from the beginning because they started the Centre for Quantum Technologies more than 12 years ago. So, it was really one of the hottest places to work in quantum computing, but also in quantum technologies altogether. So yes, absolutely. Especially in Asia, especially in this part of the world it’s one of the best places, if you're interested in quantum tech, if you want to work in quantum tech.
ML: So now, we've got just under an hour, and we’ve got a lot to cover. The audience, my audience, usually consists of people who basically know lots more than me. And also people who know relatively little, but they're all really smart. So, we're going to have to do everything from get really into some of the detail. But actually, we got to start by saying, what is quantum computing? And you know, what's the quick version please, not a PhD version?
TD: This is a question I love. It’s is a very hard question to answer. So, the quick and short and dirty answer is, quantum computing means processing information using quantum systems. Let me say a few more words about that. I guess everyone was listening in the venue is familiar with the idea of a computer, which it is what we're using right now to do this call. So, when you have a computer, what you're really doing your processing information, quantum computer goes a step beyond that, and touches on the quantum mechanical effects. So, the fundamental effects, if you like, of nature, to do something more than classical computers can do. And everyone is very excited about the concept of quantum computing today, because these devices, if you like, they bear the promise to solve some problems that are impossible to solve, or at least have to take a really, really long time to solve using traditional, or conventional classical computers, I will use these words interchangeably. And they would be able to at least, we believe, they will be able to solve these problems much, much faster.
ML: Now, is it just that they are faster? Are they kind of just like normal computers, but faster? And is it because, you know, normal computer users naughts and ones and a quantum computer can use all the different quantum levels? And there's uncertainties and some, you know, sort of weird physics effects. But fundamentally, they're just… think about them as just really, really fast computers? Or are they doing something different? And they're going to be modelling things because they are quantum computers? Can they sort of short circuit entire algorithms and do things in a much better way?
TD: I'm glad you're asking that. So no, you should not just think of quantum computers as much faster classical computers, or something like that. In fact, quantum computing is an entirely different paradigm of computation. I think there is something very, very profound about how, you know, humans learn more about nature, and learn more about physical theories, and how the computing models and the computing paradigms also improved, which is, you know, better understanding of physics. So, after having, if you like mastered classical physics, then classical computing, in a way followed suit. And as you, as I guess, you know, in the 20s, in the 30s, quantum mechanics came to be last century, and that really sparked a revolution in our understanding of nature. And in the 80s. The concept and the idea of quantum computing was born, was introduced. So, it's a different paradigm. And quantum computers can do things that classical computers cannot do, because they can tap into effects that are intrinsically quantum mechanicals, I can give you some names also for your audience, maybe they're interested in learning more, for example, they might have heard of entanglement, they might have heard of superposition, they might have heard of interference. In fact, even the basic unit of information of a quantum computing is different. It's called a quantum bit, short qubit. While the traditional unit of information for classical computers is the bit and they have a very nice analogy, that always helps me to understand the difference between the bit and the quantum bit. And hopefully, it will immediately give a little bit of understanding of why the two are different, you can think of a bit as a switch, it can only be on or off, like zero or one. And in fact, if you take planet Earth, and your position on the planet is the same as a bit, then you could only be either at the North Pole, and then you will be in a zero state, or you could only be at the South Pole, then you would be in a one state. But if you try to do the same, and you want to describe a quantum beta qubit, then you can do so much more. In fact, you could do you could be anywhere on the surface of the planet, absolutely anywhere. And that immediately should, should help you understand that quantum computers can access more computing power, if you allow me to say.
ML: That's a great analogy, I'm going to use that. Because, you know, what you've got is a situation where fundamentally with normal bits, it's binary, and you can teach it to really quite young kids using, Is there a jelly baby? Isn't there a jelly baby? You know, it's actually quite easy. But with quantum of course, it just is so non-intuitive, that it becomes very difficult to for me to envisage it and therefore to communicate it to anybody. But I'll use your analogy. So when you try and turn that into an actual thing, you know, okay, so we will read these stories about quantum entanglement and they take one particle however far away and so on. And then, you know, if you come down to the actual computers, what does it look like? What physical systems are we using? I mean, there's different approaches, right, optical, etc, etc. So, what are does it actually require in terms of, in a sense, the hardware?
TD: In fact, if you don't mind, before I answer the question, I also would like to say a few more words on why quantum computing is so fundamentally important, and why I believe it's fair to be very excited about the technology. So, in the 80s. In fact, in 1982, there was a conference organised by IBM, a conference about computing, in fact, and Richard Feynman, the famous physicist, said something incredibly profound, he highlighted the fact that nature is fundamentally quantum. And therefore, if you want to simulate nature in its entirety, ideally, you should use another quantum system to do so. But the conventional computers, the classical computers are not quantum. So, he had already understood, and he wasn't the only one. To be fair, he had already understood that classical computers have some fundamental limits, that as far as we can understand, as far as we know about physics, they are not able to efficiently simulate any other physical system. And there are also some problems that remain unsolvable for these machines. And I think it's very important to say that because it's also set the context side of the whole conversation. And I would also like to highlight how computing today is really the backbone of our economy, right? The growth in computing power, I mean, I know this is an example that everyone uses, but since the Apollo landing on the moon, and the computing power of, of the devices they use to guide, the follow up to today, to the supercomputers that we have today has been nothing short of incredible. And in fact, the more computing power we have, the more we can do as a society, the more problems we can solve, and the more we can grow all together. But there are also limits, as I said, what classical computers can do, and this is why it feels like there needs to be a transition, just like there was a transition from solving problems using you know, pen and paper. And then we introduced the analogue computer, and then the digital computer and things change radically. Now it feels like we need to do another jump, and go not throw away the whole classical computing paradigm, which is incredibly successful. In fact, keep it as it is, but also use the side quantum computers in this case. And since you were asked if,
ML: Yeah, so we're gonna get on to some of those algorithms, some of the types of calculus that we can't do with traditional computers. But let's get back to just sort of my audience has an understanding of, you know, if you walked into a room and there's a quantum computer there, what are you actually going to be looking at?
TD: And the reason why I was given this long introduction is because at the very beginning of the history of computers, these machines were huge. Today, we’re used to having a laptop. You know, on the desk, we have like two or three computers in our pockets, maybe a smartwatch, maybe one iPhone, another for work or not. So, we have computers everywhere in our house. But at the beginning, this was not the case, they were very large machines, quite expensive. They were located in few places around the world. And people would go physically there to access them, or they wouldn't, or they would from very far away, ask him to do something. Today, quantum computers are very similar in spirit, I don't have a quantum computer at home, you are unlikely to have a quantum computer at home anytime soon. In fact, there are very few quantum computers today, I would say less than 30. Or at least in that order of magnitude, let's say less than 50, commercial quantum computers of list, and they are all fairly bulky. But what is very beautiful is that unlike traditional computers, that are in fact many different ways today, you can build a quantum computer. If you go back to this idea of the qubit, which is at the end of the day a mathematical description. Then a big question that people have been asking themselves after Ferryman made that famous remark in the 80s was, how do we actually build a qubit? You mentioned a few of these architectures. So today we have superconducting chips or superconducting qubits. Qubits, have trapped ions with photonics qubits, and so forth. And if you like, we can get into the details of all these different architectures.
ML: So basically, as soon as you've got a physical thing that is exhibiting quantum behaviour, and as soon as you can trap it, keep it in the same place, and link it up to some other stuff around it, then you could build quantum computer.
TD: Very nice. Almost, we're getting there. Yes. So, a qubit is a two-level quantum system. If people are familiar with physics, we can think of that as a spin, really, but think of it as a two level quantum system. Now, not all quantum systems are two level quantum systems. So, you need to be able to find some that exhibit that kind of behavior. And as you said, you need to be able to control them exquisitely. So, in fact, the problem with quantum systems is that they're very susceptible to noise. Nature has a habit of disturbing quantum systems all the time. So, when people try to build a quantum computer, what they really want to do, they want to let me use the word create, or they want to encode qubits into physical systems, then they want to put qubits into a system that shields them from all external noise. But at the same time, since it is a computer, you need to be able to access it. So, there are always these two forces kind of fighting between each other, when you want to close your system as much as you can protect it. But on the other end, you also have to open it, because you need to interact with it and tell it what to do. Because, indeed, it is a computer.
ML: And so a lot of the challenge is you've got you'll just have some bit of noise, some bit of solar radiation, whatever it is, that comes in and creates a an extraneous a wrong answer, you have to find that and correct it within the computer so that you're not coming out with the wrong answer. Is that how… How am I doing here, because this is you're doing?
TD: Great, you're actually doing great. We are going in exactly the right direction. And I'm very happy with that. But even before we get to the what, what we just mentioned is basically the idea of error correction. So, errors are introduced into the quantum computation, caused by all sorts of effects. You want to be able to correct them faster than they actually happen. Today, we are still not really able to do that. It’s incredibly difficult to perform medical production, you need hundreds 1000s millions of qubits to perform it successfully. And maybe this is a good time in the conversation to start giving numbers and also, you know, giving names to the different players in the industry. So, what is happening today in quantum computing, you have very large players like IBM, Google, Microsoft, at some point, also Honeywell, building these devices, and you have startups, although maybe calling them startups is a little bit unfair, because they are very well funded and also fairly large sums of manpower, also building these machines, and they're all following different approaches. So, for example, IBM and Google are building what are called superconducting quantum computers. They're what you do you have a chip, and you basically build what they are called artificial atoms. So, use traditional circuitry to create something that behaves like an atom like to let like this two-level quantum system, when you cool it down at very, very low temperatures. In fact, temperatures that are lower that have the space like few million kelvins, so you were asking how does a quantum computer look like? In this case, it looks like a very big chandelier, which is just a refrigerator. And the chip is at the bottom of this big fridge, if you like, because you need to cool it down so that these microscopes or it is microscopical system starts to exhibit quantum mechanical for microscopical effects.
ML: And how many chandeliers do you need? Is that one chandelier per qubit, or many qubits under one chandelier?
TD: Many qubits under one chandelier, luckily, so there's been a lot of progress in that sense. We started with a handful of qubits. So, in 2016, IBM put the first quantum computer on the cloud. And this is another very important word that they will be using a lot today, the cloud. So, they put the first quantum chip on the cloud, it had five qubits inside the single chandalier. Today, or actually, I should say, a few weeks ago, IBM released a new device with 127 qubits still in a single chandalier, which is incredibly impressive, because in just a few years, we have witnessed what really is an exponential growth in the number of qubits.
ML: Okay, now, just for completeness, some of the startups are they really following the same… First of all, who are they? And are they following the same chandelier approach the superconducting circuit? Or are they choosing a different approach?
TD: Some of them are following the same approaches. Most of them are following different approaches. So, we are variegated. There is a company based in California, San Francisco, in Berkeley. So, they are also building superconducting quantum computers. They're also being very successful. They just announced a 40 and an 80 qubit device, still 40, 90 qubit chips. These are two different chips. Each one has its own chandelier. But then you have other companies, for example, IBMQ, the IPO just a few months ago, quite successful successfully, with a $2 billion valuation. They're building a different kind of quantum computer. And I would like to emphasize how actually beautiful this is, because in classical computing, pretty much all the devices follow. They're all the same. Allow me to say, if you buy, you know, if you buy a Mac, if you buy Lenovo, they may have different architectures. But the idea, the fundamental idea, the physical lateralization doesn't really change. This is still not true in quantum computers that are all these different competing architectures. And for example, IBMQ is building what are called trapped ions, quantum computers, where what you do, instead of building artificial atoms, you use real atoms, you eliminate an electron, so they're ionized. So, we call them ions. And then they send each single ion inside the vacuum chamber. And they keep them in place using electromagnetic fields, and lasers, really incredible, because when you look at the pictures, you can see this beautiful line of atoms perfectly aligned, again, exquisite control of quantum systems.
ML: And I'm smiling because we had Steven Chu professor, former Secretary of Energy in the US who of course, won a Nobel Prize for trapping, I believe trapping ions using a laser. And that was his innovation. But how we get them to line up and march in time with each other is incredible.
TD: And another point that I would like to emphasize is that all these ideas, everything we're discussing right now, for example, the superconducting qubits, these are based on a particular kind of qubit called transmon qubit, which is incredibly young, because the idea comes from 2007. If I'm not wrong, from Yale University, in fact, one of the people in was the author of the of that paper is today one of the leaders in the IBM efforts. So, 2007 is really, really literally yesterday, and from the idea from the theoretical idea of like the first realization of a transformed qubit. Today, we have a device that you that is basically a commercial quantum computer is absolutely fantastic.
ML: And just in terms of the chore of the different approaches, there are people working with photonics, and saying that they will be less error prone because a photon doesn't interact in the same way than ion does with all the stuff around it. Is that something that you give credence to?
TD: So right, if you remember, before, we were talking about noise, and how quantum systems are very susceptible to noise, what they say is true. So, photons tend to interact with very, very little. So that is correct percent. That photonics keeps photonic quantum computers will be more noise resistant. And there are a few companies following this approach. For example, in Canada, in California, and they're using slightly decoder following slightly different approaches, but they're both based on photonics. The difficulty there though, is that yes, they interact very little. Therefore we have less noise, but they're also more difficult to manipulate because they interact very little. So you save noise, but then you need to be incredibly smart to perform what are called quantum gates, so to perform operations on the qubits on the photonic qubits.
ML: Okay, so now let's talk about how many qubits you need to do something useful, because you're going to use some of them for error correction, and so on. So, what's the magic number? And how close are we to getting there?
TD: So let me remove from the conversation, quantum and kneelers, that are yet a different kind of beast in the universe of quantum computing. So, D-Wave is a company following that approach. They have 1000s of qubits. And I will just say for completeness, but they will not be talking about them. Because the project today follows is slightly different. What we are talking about right now is called universal quantum computation. So the finger will be saying refers to this idea of being able to perform any computation you like with a quantum computer. The largest number of qubits available today is under 27 qubit device by IBM. Then, you have Google. Okay, you have Rigetti with the qubit device, Google with between 50 to 60. There is a device in China, at the FA University that has 65 cubits if I'm not wrong, and then you keep going down. You were asking how many qubits you need to do something useful? A great question and a very, unfortunately, disappointing answer to say we don't know just yet. Because if you think of utility as worthless, if we assume the utility means being able to solve a problem that has commercial value. So a problem that somebody cares about, then a quantum computer isn't being able yet to do so quantum computer so far as not solve a problem that has commercial value that has commercial relevance, better than a classical computer can do.
ML: This is quantum supremacy, where you finally do something that you couldn't have done with a normal computer and that has not happened yet.
TD: Hmm. Right. I will be speaking about utility, exactly, to make that separation. So, there are these two concepts of sound very similar, but in fact, they are subtly different. So, what you refer to a quantum supremacy means being able to solve a problem, any problem, even some mathematical, very abstract problem better than a classical computer. And in fact, that has happened already a few times. It was not useful, scientifically maybe it deserves a Nobel Prize. And they might at some point deserve another prize, because it was an incredible effort and an incredible result. In fact, this was done by the Google's team led by John Martinez in 2019. They use a superconducting chip to solve a mathematical problem, what is called a sampling problem, where you try to sample from a probability distribution that is highly complex. And the assumption is that you wouldn't be able to sample from the distribution using a classical computer. But let's take a really long time. And when that happened, before the result was officially announced, actually, the results were leaked to the press. So, they were appearing everywhere on the internet. And at the time, I was at a conference. And there were quite a few representatives of the Google team there and everybody was like freaking out. And they were all asking what happened, what happened, what happened, what happened, and the Google team was pretending that nothing had happened. And then one week later, the results were officially published. And it was, yeah, it was it was great. But as you said, so that was a big deal. Clearly, it was a huge, it was a huge deal. Because there are still today sceptics of the technology, people who say quantum computers will never work. But I think at this point in time, they are just allow me to say being a bit delusional.
ML: So why has it not been possible to do anything useful with it?
TD: Because classical computers are incredibly good at what they do. So, if classical computers wouldn't exist, then we would be talking about a revolutionary technology, which in itself is a revolutionary technology. But the reality is quantum computers. They are the challenge is against classical computers. So you have to remember the classical computers are incredibly powerful. Like the newest Japanese device is huge. It has 500,000 teraflops. It is just an incredible, like the computing power that is machine service. Incredible. So, before we can actually reach those limits, we still have to push the boundaries of what quantum computers really can do.
ML: Okay, okay. And talk to you about how will you actually use it. It's somewhere more than 127 qubits, but it's not like it's not like we're 20 years away. I mean, it is we're getting close, right?
TD: I believe so. Yes. I don't think we are 20 years old. Yeah. Well…
ML: How will the industry be structured? You know, I want to get on to how it might have impact in, you know, the area I know, well, the net zero transition. Let's first talk about how you actually access these machines. I mean, you've got this chandelier, you're not going to have one in every, you know, in every office. It's probably if I've understood it, right, you probably need a property so difficult to handle, they're never going to end up, you know, in your mobile phone or in your car, they're going to sit somewhere, at least for the next, you know, many to they're going to arrive on the scene commercially in the next few years. But for the next many decades, they're going to be big, centralised things. What does the industry look like to enable them to be used? But it's a cloud? Is that it?
TD: Exactly. So the cloud is really the key word here, take the chandelier. And you're right, they're still fairly complex to move and be handled. But funnily enough, IBM started to ship them around the world. So, they have one in Japan, they're one in Germany, and most of them are in Yorktown in the US. So even there, there's a lot of progress. So it's becoming easier to move the machines around. But yes, for now, they either… they are sitting within the cloud infrastructure, one of these big companies, I mentioned IBM a few times, but AWS and Microsoft have also started their own quantum cloud services. For now, they're not giving access to their own device, their own proprietary devices, they've given access, for example, to get these AI infused devices, and others. So you have the central players, if you like this, quantum cloud providers, who ideally want to give you a should give you access to both the classical computing power, and also the quantum computing power. And if you're the user, let's say if you're me or my team at entropic labs, what we do is we connect through the cloud to these devices. And then we send what we call quantum programmes or quantum algorithms that can be run on the machines, and then we wait for the results to come back.
ML: Okay, so but you're not sending, I mean, it's not running. I don't know, your call centre software, it's not running your normal sort of computer programmes, because classical computing is doing all of that. But what you're doing is you're identifying a particular algorithm folding a protein or doing or doing some vast forecasts that the classical computer can't do you separate that off, and you sending just that to the quantum computer, is that correct?
TD: This is one way to think about it. Yes. And you very correctly, you said something very important, right, which is you cannot just take your code, let's call it your classical code, whatever code you might have, send it to a quantum computer and hope that it will, they will work because it is not going to work. The paradigm is different, the logic is different. In fact, even when I speak about our company it’s very easy to say that we are a quantum software company, because we don't build quantum hardware. But I'm very uncomfortable with that word software, because I personally believe that software in quantum computing is still mostly lacking. But it's lacking simply because we never needed it. Because we didn't have quantum computers on the cloud. So, this is a whole new industry that is growing very, very quickly. And we are starting to realize what is missing in terms of hardware, infrastructure, and software needs, and people are filling gaps, etc, etc. And to give more meat to what you were saying, like we work in optimization. So applied mathematics, ideally, you have this complex problems, you want to find the ideal configuration that either minimizes or maximizes some cost function of interest for you. So what you want to do, you want to find the mathematical description of the problem, you want to rewrite it so that you can encode it into the logic of the quantum computer. And then you interface with the quantum computer of your choice. Remember that we have different devices available. So there is still quite a lot of work needed to interface with the different machines, then you send the code, and you wait for the result of mine. So
ML: I think when we first spoke, or one of our early conversations you talked about a typical algorithm might be inverting a matrix, something that's incredibly computationally intense, and might take a classical computer, you know, if it had a big enough matrix to take 1000s of years, so you just send that to the quantum computer, then you get the results back and you kind of then go off and continue doing whatever it was optimizing your power network or whatever using the results.
TD: So that could be one way to do it. And I'm emphasising the fact that it could be one way to do it because the reality is that people are proposing different approaches. as to how we will be using quantum computers. So, inverting a matrix is something that becomes very, very important when you want to solve linear systems, for example. So, in engineering in data science and machine learning, this is a fundamental problem. And yes, it can be very complex if the matrix is very large. So, you could imagine in the future, and I'm saying in the future, because to solve that problem you need, most likely millions of qubits. So, you're still very, very far away from being able to handle it in practice, what you could do, you could have a workflow, where most of the computation actually happens on a classical computer. And then once you have the matrix, once you have the information, you pass that to the quantum computer, using a particular architecture, they're still missing colour quantum RAM, quantum random access memory, you do what you need to do, you extract a piece of information of interest from that computation, you return it to the classical computer, and you continue with your workflow. That could be one way to do it. If you are thinking about quantum simulations, for example, quantum chemistry, material science, then it becomes even more complex, because it's still unclear how much of that computation is to run on a classical computer, much of that computation has to run on a quantum computer, because again, these are hard regimes to explore.
ML: So you may actually have a model of a particular of a molecule or molecular interactions in the quantum in the quantum computer. So then now, question I'd have there though, is, how do you check your maths, because you send something off, and it comes back, and it says, Write this, this drug will interact and we'll do this. And you say, Well, how do you know? And it'll say, well, because I'm a quantum computer, and you have no way of knowing whether that is in fact the case? Or do you? Is there a way of sort of, I don't know, using other quantum computers to check each other's maths?
TD: It turns out, this is actually a huge problem. Because if a quantum computer can solve problems that a classical computer cannot solve, then if a quantum computer in the future gives you answer, you cannot rely on your trustworthy classical computer to check it, because it cannot derive the same answer. So, what do you do, there's actually a problem called verification of quantum computation. And if you assume, and let's say it's a fair assumption that quantum mechanics is correct, then what you want to ensure is that the quantum computer is actually doing what you're asking it to do. Or that the quantum cloud provider is, indeed, giving you access to a quantum computer and not to some bogus machine of source. So yeah, you mentioned something that reminds me very much of a work… I was part of a collaboration that was part of a few that was published a few months ago, where indeed, we use different quantum computers to check each other. Now that we have more quantum computers available, then it seems quite a reasonable idea, to use each device to check each other, and, you know, convince ourselves that they are actually doing what they should be doing.
ML: I'm just reminded of a time when I think it was one of the Intel chips did a particular thing. And I don't know if it rounded down or it did something and it actually came up with wrong answers. And it took a long time to figure out what was going on. I don't know if you're familiar with this situation like that on steroids.
TD: Yes, absolutely. Because that is an even more subtle problem. But here we are really talking about devices that, unfortunately, are still suffering from all sorts of noises. So, all different issues, if you like, so yes, you can say that this is something very similar on steroid.
ML: Okay, so now, now with that all as the preamble and the caveat, how hard it is and where we are, and so on the transition to net zero, there are all sorts of problems that I could imagine might yield to this phenomenal new computing power, in whether it's within the chemistry or the network modelling are all sorts of things. Are there any that you would sort of pull out and say, Yes, I could see how quantum computer is going to revolutionize this or that aspect of the net zero transition, I don't know, anything come to mind, or I could suggest a few?
TD: Quite a few actually come to mind. And that's another reason why I am personally very excited about quantum computing as a technology. Maybe we can start with this division, but we can categorise problems into two set of problems, problems that are very natural to quantum computers. So let me say that they are grounded on concepts of quantum physics and quantum mechanics and problems that are not quantum per se. But they rely on basically very complex systems. So, you're trying to solve applied mathematics problem with a very high degree of complexity. So, in the first category in the former category, you would have quantum chemistry, molecular simulation material science. And those of as you easily imagine, a lot of applications into the…
ML: Right so things like enzyme reactions, battery chemistry, catalysts, or cap in the battery cathode. Exactly. So surface effects in photovoltaic. Absolutely, there's something quantum going on. And therefore trying to break it down into a linear classical computing ones and noughts is just simply not going to capture all the things that are going on at the same time.
TD: Correct. So, if you want to fully understand the properties of quantum systems, intrinsically, you need a quantum computer, you need another quantum system to capture those properties fully. And you mentioned fantastic use cases, right? Especially for me something that I find very, very powerful this idea of batteries. Now we're witnessing this transition, if you like to having incredibly good, let me just say the word good, incredibly good batteries. And this is one of the reasons why we're also witnessing this transition to electric vehicles. Because now we can actually keep a car with a battery that lasts long enough, that can be that can be quickly recharged. And before this was not possible. So, imagine a future where we can improve on that even further, where we can actually increase the capacity of the battery. But we can increase the lifespan of a battery, where we can start miniaturise the battery that enables not only it helps with reducing, you know, emissions, and it helps with all of climate change. But it also enables new technologies that today are just hard, especially in transportation, especially in production, and also in grid management of the energy grid.
ML: And it's a sort of dirty secret in some of those sectors, how much trial and error is still required. Just formulating different bits of the battery or, or different combinations. It's the same in the in the drugs industry isn't the same in the medical industry, there's much more trial and error than you're led to believe by watching Hollywood films, right?
TD: Yeah, absolutely. So that problem is that in chemistry and in material science, the worst cases in simulation may require exponential resources in terms of computing power, which means that if you want to fully simulate the system, it will take too long, just potentially 1000s if not longer of use. And you want exactly to avoid the problem that you're mentioning, trial and errors, lab train and arrows, you want to move as much as you can of that process, let's say on silicon and to schism quantum computation, so that then you can only go and test what you're already fairly sure it's going to work.
ML: And are we already seeing companies trying to do this? I mean, the computers aren't quite ready for primetime. But are we already seeing the big chemicals companies that the Dow's and, and the BASF's and so on? I mean, I don't want you to reveal who your clients are. But do they include or do you see the big chemical companies, the big material science companies, you know, getting actively involved now in, in quantum computing?
TD: Right, so? Absolutely. And you see, in fact, it's very interesting. You see more than that, you see the quantum hardware companies, you see the startups, and you see, the chemical companies say the big companies interested in these applications, all working together, it's still a very collaborative ecosystem, I would say. So you're, for example, Microsoft Station Q, which is a research centre, sponsored by Microsoft in California, they've been doing incredible work in 2017 on nitrogen fixation. Okay, just to give you an example…
ML: Nitrogen fixation, of course, is a really a huge problem to solve for the climate, because at the moment, we use nitrogen and fertilizers. And I think it's 2% of global emissions comes from making fertilizers from natural gas and even from coal, so, we make the hydrogen even from coal. So, fixing nitrogen from the air in a clever way would be huge.
TD: Right. So, exactly, so I think it’s about between 1% or 2% of energy consumption globally, it goes to solve for nitrogen fixation, because you want to have nitrogen basically assume a biological processable form, using catalysts. And what is incredibly fascinating about this problem is nature knows how to solve it. Because plants bacteria they do that is just it is incredibly hard to find the right catalysts. So catalysts are very efficient for our needs. Because it's just a hard problem. There are too many you can practically test all of them. So you eventually if you want to improve the processes, it is not today You need to rely on computation in the in this case, quantum computing is going to help.
ML: So the question of nitrogen fixation, of course, there's a parallel problem around carbon fixation of carbon out of the atmosphere in an incredibly elegant and clever way. I mean, the thermodynamics of carbon is very different from thermodynamics of taking nitrogen out, because it's 400 parts per million versus 80% of the air or whatever, whatever nitrogen is. But that feels like an analogous problem.
TD: And funnily enough, Microsoft is also looking at that the release of it paper just recently, where they are talking about CO2 fixation. And again, the idea is very analogous, right? They are looking for quantum algorithms so that they can find better catalysts for CO2 fixation. Also, in terms of carbon capture, since you mentioned that, there was a recent collaboration announced by Total, together with a startup called at the time was called Cambridge quantum computing. Today, it merged with Honeywell quantum solutions, and they're called Quantum. But this is just to say that is not only Microsoft does, not only big players, but as you were mentioning, is also the big companies together with startups looking at innovative solutions for these problems using quantum computing.
ML: So those are examples of processes that are inherently quantum No, it's materials, its chemicals, it’s batteries, it’s surface interactions on solar cells. And those ones, you reckon you've kind of got to go quantum because they just the nature of the problem, but you said that there was another one, where you're going to use the quantum computer almost as a kind of bigger Sledgehammer for problems that are fundamentally, you know, accessible to binary processing. But it would just take too long or be too difficult. Can you give some examples of those?
TD: Sure. Yeah, exactly. I think of that as applied mathematics, let's say again, on steroids. So, I think the key word there is a complexity. So, these are intrinsically very complex problems. And they happen in optimization, which is the field we were focused on, but also in data analysis, and machine learning, artificial intelligence, and fluid dynamics, or weather forecasting, and so forth. So, I'll give an example that I think is incredibly interesting. Speaking about transition to renewable energies. Now that the price is both of basically, solar production and wind production have been quite amazingly collapsing. Pretty much an exponential rate in the in the in the last 10 years by I believe, between 80% to 90%, at least for solar cells, is not crazy to imagine that in the near term, we're going to see big changes in the electric grid in the way the electric grid is designed. In fact, this was something that Noah Smith recently discussed also in his blog. If you if not the blog, you should follow it is resist. So you can you can imagine that you're going to have an electric grid, this is going to be very different from the one that we have today, potentially, where there will be a lot of smaller producers. But then having a big coal plant in the middle that is giving electricity to all the nodes, each node might be producing electricity independently. But if these nodes are producing electricity based on renewable energies of renewable sources, so then you have stochasticity in the process, because you know, if it's cloudy, if there's no wind, you're not going to be able to produce electricity on the time. So, you need to be able to optimize this huge problem, where you want to have a flow of basically, you want to have power on the full grid constantly, because everyone wants to have electricity available right at each point in time. But you need to do so by basically optimizing over a very large number of points, each one producing electricity in a stochastic manner. This is a huge optimization problem. Now, classical computers might be able to handle it, we don't know. But it seems to be quite a nice problem for quantum computing. In this case,
ML: Now, we could disappear down this rabbit hole because there is an alternative approach there. And you know, I love that, you know, applying his fabulous mind to this stuff, but there is an alternative approach, which is to have the nodes talk to each other and react like a flock of birds to what's happening around them, rather than having a single central super control, which would have to be more powerful than any computer that already exists. And, you know, I was, you know, this would be the transactive grid, which, you know, I don't want to sort of pull rank but I've been trying to get my head around for Since about 2008, and you see elements of it, and the answer is going to be, you're going to see elements of both, you're going to see the trend, probably the bigger grid, the transmission grid, centrally controlled, and the distribution grids. But when it gets out to the edges, you have to have the washing machine talking to the to the other appliances in the house, not having that solved by even a quantum computer in Houston, and then trying to send the message back through the system, I think is going to be very difficult.
TD: You're absolutely right, it’s is one of the reasons why I love optimization. Because sometimes we're able to tackle very complex problems using various Mercury sticks, for example, instead of looking at the global problem might be enough to look at the local problem. So break it down to local problems and find good solutions. But as I think what you said is very real, you have edge problems, but you also have global problems that more likely than not, are going to require the high level of computing power to be tackled.
ML: Absolutely. And I don't want to push back against the concept of the quantum computer in the optimization role, then another one that would come to mind clearly would be the transportation system, where you actually have quite a lot of central control, and communication, even optimizing London's traffic light system. If I told you how primitive it is, today, I was on the board of Transport for London. And I thought that there was some magic, it's not magic, it's still a lot of engineers going around and not quite restock watchers. But it's certainly not being done in you know, in a way that a quantum computer in theory could sit across it and do.
TD: I'll give you since we're speaking about transportation, I'll give you another interesting problem there, which is the whole idea of electric vehicles, grids optimization. So, if you have a limited number of stations, for example, but you can charge your EV, and you have a very large number of electric vehicles moving around, you want to be able to find, first of all, you want to be able to position the charging points in the right places. So this is a design, if you like, problem at the beginning is like a planning problem. And then you have the scheduling problem, which I believe is what you were referring to, like once the vehicles are moving, you might want to have something telling you where to go to charge the vehicle for long, etc, etc. So, it's just I find it incredibly interesting to think about this very complex optimization problems very quickly pop up pretty much everywhere, once you start to have a very large grid with a lot of moving nodes, and with a lot of stochasticity. So with a lot of, let's say randomness intrinsic. In fact, there's I think it's also important to mention that there is there is an organisation or consortium I think is the right word called Q4Climate, which brings together different scientists and experts, both from startups and universities and academic institutions, that is also looking at this kind of problems, you know, in chemistry, in optimization in also in quantum sensing, for example.
ML: Well, that's fabulous. And what we'll do is we'll put a link into the show notes because I wasn't aware of Q4Climate and it sounds like I should have been, but I wasn't. So let's try and diffuse that knowledge as well. I got to ask you this. We've been through the kind of you know, there's the sunlit uplands, we will be solving chemical problems and quantum problems, we'll be doing this huge Sledgehammer optimization problems. But is there a concern that the first quantum computer will actually be used just to crack the code and break the security on all of the infrastructure, all of the energy and transportation passwords out there? So, you know, is there a risk that the first impact of quantum computing on the infrastructure and energy transport climate will be massively negative huge amount of investment required to protect against somebody using quantum computing?
TD: Well, there certainly is a risk. In five, you're referring to referring to the infamous and famous Shor's algorithm, though this is from 1995 is one of the very first maybe the first quantum algorithm that had a real world application, unfortunately, not necessarily a, quote unquote, good one. And this quantum disk quantum algorithm can be used on a computer on a quantum computer of sufficient size. And again, we're speaking of millions of qubits, so something that is still fairly far away in the future to crack RSA and RSA, which is the encryption algorithm. It’s like a public key distribution. There's used ubiquitously on the internet today. So usually, when you connect to your bank or to your email account, you are communicating securely with the server through a variation of RSA.
Right or to your power station or to your goal or to your to transmission grid, your high voltage grid, I mean, it's being used to secure everything.
TD: Correct. Pretty much everything now, yes, there is a there is a threat, let's call it is quantum threat that if somebody somewhere right now at a quantum computer of sufficient power, yeah, they could crack all this codes. And it will be very hard for us to know that this is happening if they're smarter about it. I don't believe this is happening. And in fact, I believe that the knowledge that this is the really this was the first application, or the first well known application of quantum computing this knowledge, in fact, I think it has been way more positive than it may be negative in the future. And let me say why it because it push the field forward. So it transformed quantum computing to what was at the time, it really was more of some sort of curiosity, some sort of computer science niche topic that people were looking at. But once people became aware that it actually had a huge application, this completely changed the game. And companies, governments started to pour more and more money up until today. So the fact that everyone is aware that this is an application of quantum computing, I actually think is a very positive thing. First of all, because it forces everyone to be very open and transparent about what you're doing. But also because it forces companies and standard makers to prepare for a threat. And everybody knows about it.
So every digital technology that takes off, has applications in one of two areas. One is security. And the other, of course, is pornography, but I don't see quantum computing being much use there. And maybe I'm wrong, sometimes some clever soul will come up with something, but the point about it being a potential threat, and that focuses the mind. So that has sucked in a lot of resources. Every presumably every CTO, CIO, chief information officer had to have a view as to whether it was going to render security inoperable. But does that also mean that there's some enormous arms race between not just companies but also between countries for quantum superiority? Not over classical computing, but using the words in a different way? superiority over each other?
TD: Sure, there is. Yeah, yeah, there is. And it's also important to keep in mind the quantum, which is what we are discussing, but quantum also gives. So yes, there is a threat. But, you know, governments countries are aware of that. And there certainly is, let's call it an arms race between the US, Europe and China, where they're all rushing to build more and more powerful quantum computers before the other countries of blocks do. But there is also an effort in building new infrastructures based on quantum communication, for example, also, based on quantum computing, where you are able to perform certain cryptographic protocols, certain security protocols in a way that you wouldn't be able to do with classical, either communication or classical computers. With much higher security, if not complete security, complete security. And cryptography is always a very theoretical concept. That doesn't hold in reality. But I really love this concept the quantum takes. But also quantum gives, like quantum key distribution is an idea similar to RSA in spirit, based on entanglement, the effect that we mentioned briefly at the beginning of the conversation, that allows you to exchange a key between two parties in a way that is basically unbreakable in the sense that somebody is Eavesdropping is listening to this key exchange, the two parties will be able to detect it, and immediately stop the communication.
ML: And this is presumably to do with Heisenberg and all sorts of clever quantum physics that if somebody listens in it changes the nature of the algorithm of the calculation. Now it's
TD: So it has to do with the collapse of the wavefunction, he also has to do it really is based on on the very funky properties of entanglement that allows you to do something that you can do with classical correlations.
ML: Very cool. I need to stop there, because you're going to lose me. But we're also nearly out of time, Tom, and it's been absolutely fascinating. I want you to just finish if you might, if we were having this conversation, and hopefully we will be in five years time. Can you tell me what we might have seen? What would have been what what is it that you think is so close to happening that in five years time, you'll be able to say, Michael, I told you this would happen and it did.
TD: I hope that it will be quantum advantage. What you were asking me about before the demonstration that a quantum computer solve a problem of value, whatever value means, in a way that a classical computer is unable to do. I think we are very close to that. That the biggest transition that we have, that we have enjoyed in the last really and he has his that his machines moved from being, again, university curiosities. Curiosity is an exaggeration, university endeavors to commercial devices that anyone can access that a teenager can access and start playing around with. And it is. So we are really living this stage or this time in the industry, where people are trying all sorts of stuff is the stage of quantum computing, if you like, and I believe that there is a very high probability, there's a real chance that by having access to these machines by pushing the limits of what they can do, by trying smart and clever algorithms, we will be able to find a niche application where quantum computers can exhibit quantum advantage, whether it's going to happen in the next five years. Well, we don't know, nobody really knows. But if you asked me Could it happened. And in five years, we will be talking about that. And we will be very excited. Yes, I think it can happen.
ML: Very good. So here's hoping that not only will quantum computing have solved, its first real world problem, but that that solution will actually also help us move towards Net Zero, rather than some other, less vital aspect of the economy.
TD: Absolutely. Very good.
ML: Listen, it's been a great pleasure talking to you, Tom, thank you very much for spending some time with me. I know it's very late over in Singapore. And so, I'm going to thank you on behalf of all of our listeners and our viewers, and go. Thank you.
TD: It has been a great pleasure to think this much Michael.
ML: So that was Tomasso Demarie, a CEO and Co-Founder of Entropica Labs, and quite the expert on quantum computing. My guest next week is Inger Andersen. She's an Undersecretary General of the United Nations, and she's the executive director of the United Nations Environment Programme. Please join me at this time next week for a conversation with Ingar Anderson. Cleaning Up is brought to you by the Liebreich Foundation and the Gilardini Foundation.