Transcript | Quantum Computing Standardisation— with Pasqal

Standardisation is something we take for granted in classical information technology. But when it comes to quantum computing, are we all speaking the same language? How is one system manufacturer trying to ensure that all aspects of QIS will interoperate going forward? Find out how in this episode of The Post-Quantum World. I’m your host, Konstantinos Karagiannis. I lead Quantum Computing Services at Protiviti, where we’re helping companies prepare for the benefits and threats of this exploding field. I hope you’ll join each episode as we explore the technology and business impacts of this post-quantum era.

Our guest today is the vice president of global policy and partnerships at PASQAL, Catherine Lefebvre. Welcome to the show. 
 

Hello. Thank you for having me.

I see you have Ph.D.’s in both molecular physics and theoretical chemistry and quite an interesting background. We should definitely start with how you found your way to QIS and focusing on this field.

I developed this passion for quantum in 2000 when I was an undergrad student. Twenty-three years ago, I started diving into the Schrödinger equation and thought it was pretty cool. I did from my undergrad to master and Ph.D. I love understanding nature at its microscopic level, so I studied laser-matter interaction, was doing simulation, solving the time-dependent Schrödinger equation on a simulator — using HPC resource, the type of simulation you would use hopefully one day on a quantum computer. But back in the day, we didn’t have that, and still, today, we don’t have that for this kind of simulation. But that was looking at ultrafast dynamics of correlated electrons and nuclei in small molecules, and it was a lot of fun.

I joined the thesis program between France and Quebec. I’m Canadian, and I studied in France, very close to where PASQAL has its office now. And then I continued. My passion for research and quantum continued, and I decided to do postdocs. And I did three postdocs — did research for another nine years or so, in the same type of field: laser-matter interaction and ultrafast dynamics.
 

What I’m hearing is, very early on, you were able to go from that theoretical — let’s say, the Schrödinger equation, and it sounds so fantastical and abstract — and then you were able to bring it to the lab and experience some aspect of it. That’s exciting — that first time that you could bring it to life. That’s what quantum does in general — it’s letting us touch that other layer of reality.

Indeed. I was in what we call the quantum Revolution 1.0, where we were studying quantum phenomena, observing quantum phenomena, but not at the point of engineering these processes.

So now, bringing that practical aspect to PASQAL, let’s talk about what PASQAL does. Can you give our listeners a high-level intro to the neutral-atom technology?

PASQAL is a neutral-atom quantum computing company, full stack. We develop the hardware, middleware and software application, algorithm and so on. And on the hardware side, it’s again, laser-matter interaction. We’re using lasers to control atoms, to trap them and build a matrix of atoms and then encode information and process information in a very controlled manner using lasers. And neutral-atom has quite interesting advantages as compared to other modalities.

Each of the modalities in quantum computing at this early stage of the technology has its pros and cons, and I’m going to just explain the pros of neutral-atom. The fact that we’re using atoms as our qubits, the qubits are the most pristine and perfect because there’s no fabrication, and the atoms are built by nature. Intrinsically, it is perfect, so that has multiple advantages. And then you have all the flexibility where you can have a 1D, 2D or 3D matrix — I shouldn’t say matrix for 1D — but you have a flexibility in the geometry of your qubits and programmability using also laser pulses to encode the information and play with that.

And then, in terms of scalability, it is easy to add more atoms and build qubits out of them without any fabrication steps. And it’s still in the same device, which is not that big. It fits in a relatively small box — nothing to compare to the big dilution refrigerator that you need for superconducting qubits, for instance. Neutral atoms are in this — you have a very small vacuum chamber and all the other elements around it, but it’s still small, and it is at room temperature. That’s a huge advantage. It uses much less energy compared to other modalities.

Another interesting thing with the neutral-atom is that you can operate in two modes — analog mode or digital mode. The digital mode is the one that will deliver the universal quantum computing and all the power of it at fault-tolerant quantum computing. But we’re not there yet. No one is there yet. And we’re developing this alongside the analog mode, which is, for now, what we see as the most promising to achieve the practical quantum advantage early stage in the next few years for specific applications. At PASQAL, we’re looking at the two modes at the same time with the very same device, but we can achieve both. 
 

We have a credit risk offering based on a paper written using a PASQAL system, so they’re obviously real. We know that. We have evidence. How can our listeners access these machines today?

The best way to access it is to partner with PASQAL. That’s what we’ve done. That’s a big, nice demonstration that we did that was also a collaboration with Crédit Agricole and Multiverse Computing. But we have a different level of offerings, and the first level is what we call quantum discovery. It’s to onboard new customers and partners so that they learn not only the quantum computing itself but also the specificity of neutral atoms. And they start coding. They start also testing use cases. That brings them to the next level of defining targeted use cases and starting proof of concept. We have different levels where we work with customers.

Also, we are finalising the integration of the device, the emulator, and the quantum processing processor on Microsoft Azure Quantum, and it should be available, then, through the cloud for anyone who wants to create an account on Microsoft and then use it. That’s two ways. We’re also working with high-performance computing (HPC) centers so that we are fully integrating our quantum accelerator in their HPC system and offering. There’ll be one installed in France and one in Germany. HPC users also, as part of that ecosystem, can eventually use it. But today, the best way is to partner with us, and then we develop the best offer for each partner. 
 

And you mentioned that you do the whole stack, so you could set it up however you need to that way.

Yeah. Some are interested in just the hardware integrating into their HPC offering — you add a GPU offering into your HPC system. And then some others are definitely not interested — not even in coding — so we are defining the use case with them and then defining the full method and the solution, and then implementing eventually into that workflow.

It sounds like a box that might end up in data centers very easily. It’s not, like you said, a gigantic refrigerator or whatever. It could, in theory, be a rack, like a 2U rack or something.

That’s exactly what we are building.

That’s exciting. Can you describe your role at the company?

My role at PASQAL is global. I’m in charge of the academic partnerships — all the research partnerships we do across the globe. And we do that to accelerate our technology roadmap and find experts all around the world to help us de-risk some parts and advance faster. We do that on both the hardware side and the algorithm side.

Another piece of my role is standardisation. I’m setting up the strategy for the company on quantum computing standards. I have multiple hats there. I’m chairing two working groups at IEEE, and I’m convener of the new joint technical committee at the European level. I’m convener of the working group on quantum computing, and I follow what happens at ISO and other standards bodies.

Aside from that, I make sure that we’re part of the ecosystem. The company is headquartered in Paris and France, but I’m here in Boston in the U.S. And I’m Canadian, so I have strong roots in Quebec and I know the ecosystem, so I’m making sure that we are part of the North American ecosystem and being part of building the community. I take part in all the activities at the QEDC and locally in Boston, working with different startups and other players or actors in the ecosystem so that we have a stronger and more cohesive hub in the Boston area. 
 

It’s amazing I was able to get you for time to do this recording.

It’s my pleasure.

Let’s talk about the importance of standardisation. In quantum computing, it’s one of those words that you hear about across the industry, and people forget about it after something becomes commonplace. So let our listeners know some of the benefits and concerns around standardisation.

I hear a lot of people saying it’s way too early to start standardising this technology that is still under development, but I’d rather we start early and define a common language so that once the technology is mature enough, we already speak the same language and we can define the concrete standards. I like what NIST is putting together, mapping the standard-level readiness with the technology-level readiness levels of TRL, with their SRL. And if we look at quantum computing, we are at the level of commercial prototypes.

We’re not in implementation in the field yet at scale, but we’re still pretty advanced. And if you’re looking at the scale that NIST has put together, we’re already defining the terminology of quantum computing, the vocabulary: What is a qubit? What is a quantum computer? What is a quantum simulator? What is a simulation? There is a huge confusion, and when I go to Europe or I speak to people in the U.S., we don’t have the same definition of simulation.

This is already the time to set what is what, and then we can define the functional description of different pieces of a quantum computer: What are the commonalities between the different modalities? The different superconducting qubits do not operate the same way as neutral-atom qubits, and we don’t use the same protocols the same way. How do we benchmark? Should we benchmark against the hardware, or against the application? There are debates like this to discuss now with the experts. The benefits — first is to speak the same language amongst the community. And that’s an effort that should be international, that should be with all the experts in the field. And you see already some interesting play between different countries or different experts.

For us, PASQAL, we’re developing neutral-atom quantum computers. That is not the mainstream, so we want to make sure that if standards are adopted and they do not reflect the type of technology that we develop, we don’t want to be forced to adopt a standard that does not correspond to the technology we have. A standard eventually is a stamp, a proof, that we were delivering something interesting for the market. And if we cannot have proof that, yes, this is a real quantum computer, or this functions this way and this way, and we cannot benchmark against something that is adopted in the community, it’s not helping us. We’re making sure that our technology is well represented while it is so early-stage in the development of standards. That’s why we’re part of this effort at the European level and at IEEE. 
 

I wanted to draw some attention to this because when people think of NIST, they just think about the standards for post-quantum cryptography. And next year, we’re going to have standards for protecting against the threat. That’s 10 years away, but we still haven’t figured out what these machines are when it comes to standardisation.

I was working a little bit in standards many years ago in AI, and I was on the ISO’s mirror  committee in Canada. At that time, we had five definitions of what artificial intelligence is. In 2018, we didn’t inquire yet how to define AI. Today, we’re trying to urge people that we need to regulate on AI and the application — better to start early and agree on definitions and terminologies and benchmarking methods and metrics and so on. Otherwise, we’ll try to catch up. And standardisation is a very slow process, and the technology has evolved very quickly, so we need to react quickly in terms of standards.

It’s a good time to shift to a little AI discussion here. How does your background on developing AI ecosystems come in handy in the quantum realm?

I’ll tell you how I ended up in AI and how I ended up back in quantum. After being a researcher for almost 14 years, if I include my Ph.D. studies, I decided I have this passion for quantum, for research, but I also have this passion for collaboration, and mostly global collaboration. It started from the time I was doing my Ph.D. between Canada and France, and I was leading this big research consortium with experimentalists and trying to find common language between theorists — my group and the rest of the experimentalists. And bridging between physics and chemistry and the different topics and different countries, I enjoy that. To me, this is the way: Bringing experts together, we can advance knowledge and we can help the development of science and technology.

This is how I carved my career after being a researcher. At that time, there were not so many opportunities in quantum technology — it was still mostly quantum science — so I had some opportunities and surfed the AI wave. I worked at a bank, and I worked at a startup. I was building bridges between academia and the industry and with government. I was making sure that we bring experts to develop some application for the bank, for that startup, and translating to the business what the research was doing and what it meant. And then I had a short contract with the Quebec government as the adviser on quantum technology, and the Quebec government is central to building the local ecosystem in quantum.

It’s through these experiences that I created this: How do we work with the ecosystem, and how can we make them strengthen the ecosystem in AI, in quantum? But I have to be honest that my passion has always been in quantum. Even when I was working in AI, I kept following what was going on in the quantum sphere and how the industry was growing. To me, the role at PASQAL is perfect because I’m bringing all the aspects that I like, and it’s global. And we see a lot of positive impact already. 
 

I get it. I hope this isn’t some kind of deep winter, like we experienced with AI years ago. I definitely don’t want to see that. But AI has stolen a little bit of quantum’s buzz this year. People didn’t expect it. No one expected what would happen in late 2022. Do you think, though, that there’s an opportunity here that quantum ML or some other synergistic use case might take advantage of this AI wave and bring quantum back into the spotlight?

I don’t think they are competing. Let me say the quantum winter is still a debate, and maybe with global warming, winters are not too cold anymore.

That’s a great point.

Quantum and AI are not competing, and we’re seeing a lot of development in quantum machine learning, for instance. And if I go back to this use case that we did with Crédit Agricole and Multiverse, this is a quantum machine learning use case where we developed this credit scoring application detecting Fallen Angel. And what we observed in this first implementation on the hardware of Qboost, a new method we developed, is that with a much simpler and faster method, we reach the same performance as Random Forest, which is the state-of-the-art machine learning method. It’s not quantum advantage at this implementation, and that was only on 50 qubits of our device. But then, if we do a simulation with more like 90 qubits and a bit more flexibility in the qubit interaction, we can reach quantum advantage, outperforming the classical method for a useful task, for a business case.

It is something that is on our roadmap that in the next year, we can most likely achieve. That means bringing quantum and machine learning together for something that is useful. And that was convincing enough that Crédit Agricole is putting this solution into their workflow already.

At PASQAL, we’re also working on graphs, and it’s interesting because you can map a problem onto the geometry of the atoms — map one-to one or natively the graph problem. You can move the atoms and then map the same type of graph so that the problem you want to solve, you can solve in a lot fewer operations, so it’s much faster. We did that, and then much faster, but then we also have less error that you accumulate. We are doing that on a drug-discovery problem and looking at using a real data set on a biological difference system. Again, we benchmark against the state-of-the-art graph machine learning classical method and arrive at the similar on par performance — not quantum advantage today, but promising that we can achieve that very quickly.

I hope that answers the question of AI and quantum. I don’t think they’re competing, and they’re working together. This is not the dream of the quantum simulation of chemistry, of solving the time-dependent Schrödinger equation for this big molecule. We’re not there yet, but there are so many other problems that are interesting and that can bring a lot of business value that we can start working on today. 
 

I view it like a competition only because if you think about research-and-development budgets, some companies are thinking along the lines of “We’re going to spend some R&D on something this year. It’s got to be AI, because that’s what everyone’s talking about.” You could see that happening. When they’re pulled together into one, it makes a very compelling case.

I see that.

What other real-world applications for quantum technology show the most promise for helping quantum show its advantage as soon as possible?

At PASQAL, we are working with different industry sectors, different verticals, working in the utility vertical with optimisation of grid management, working on the automotive vertical — for instance, with BMW, we a good partnership, looking at the new batteries and the deformation of metal when there’s a car crash, for instance. We’re working on a climate simulation with BASF. And in pharma, we developed a new method for drug discovery, a neuro-optimisation quantum external learning method.

I’d love to hear some thoughts about the roadmap that PASQAL has planned for the machines — where you see it going, qubit counts, the switchover to gates for digital.

What I can share is that we’re, very short-term, pushing the analog mode and the analog/digital mode, the hybrid, because this is where we see we can achieve quantum advantage first. We think that the ecosystem needs good demonstrations that we can achieve quantum advantage. Hopefully, PASQAL does it first. The whole ecosystem needs that, so we’re working on that. Our device has 100 qubits — a commercial device. We have a second generation of 200 qubits. In the lab, we’ve achieved 350 atoms. We know that we can scale with the very same device up to 1,000 qubits without any fabrication or any change of the device. And with the same device, we can also develop the digital mode. It’s a synergic and continuous development, both analog and digital.

Do you view that about 1,000 would be the ideal number of qubits to shoot for in that device?

There will need to be a modification of the device to go beyond that number of qubits.

Maybe higher-dimensional matrixes or something — different shapes?

Yeah. But there’s no need for interconnect. It’s just the same. It’s not in a chip where you need to connect different chips together as we scale. It’s all the same type of device.

The current roadmap is to keep scaling by that method — dimension and size — and just keeping it all in one kind of chip?

Yeah, for the next one to three years.

That seems realistic.

You mentioned winter, and it’s getting a little tough on the quantum side. But it is the case for all deep tech in general, from what I’m hearing from investors and analysts. But at the same time, at PASQAL, we raised our Series B of €100 million this year, and this is the biggest raising in Europe so far. We’re still seeing a lot of traction. Also, we’re working with partners already and building capacity internally. It is the way the industry needs to go. Even if they’re not reaching quantum advantage tomorrow, tomorrow plus one, they’ll be prepared and they’ll have capacity internally, and that will bring them a huge strategic advantage against other competitors. The fact that we’re working with the industry already, and since the beginning of PASQAL, that helps also.

What’s exciting and impresses me is how many of your use cases are knocking on the door of advantage. They’re right there — a few percentage points off, if you want to consider it that way. It’s going to be exciting to monitor this. Thanks so much for sharing all the information about your company and the approach. I appreciate it.

Thank you very much for having me.

Now, it’s time for Coherence, the quantum executive summary, where I take a moment to highlight some of the business impacts we discussed today in case things got too nerdy at times. Let’s recap.

PASQAL makes neutral-atom quantum computers that operate at room temperature and use less energy compared to some other modalities. The neutral atoms are controlled by laser and should allow for large numbers of qubits on one device. Organisations can experiment with PASQAL systems directly through the company today, but expect to see access via Microsoft Azure Quantum soon.

In addition to helping PASQAL accelerate its roadmap, Catherine is involved in standardisation efforts in quantum computing. She chairs two working groups at IEEE and is involved with the new Joint Technical Committee in Europe. She closely follows what goes on at ISO and other standards bodies. The goal is to ensure that we have a common language in this early industry stage, and it’s what we should be doing, according to the NIST ranking of our industry as being in the range of commercial prototypes. Next, we’ll need to tackle the different modalities of qubit types and how we benchmark such different entities.
Catherine’s pretty experienced at performing this sort of role. In 2018, she tackled similar concerns in the growing field of AI. PASQAL is now working on quantum ML use cases, so all this comes full circle for her.

That does it for this episode. Thanks to Catherine Lefebvre for joining to discuss PASQAL, and thank you for listening. If you enjoyed the show, please subscribe to Protiviti’s The Post-Quantum World, and leave a review to help others find us. Be sure to follow me on all socials @KonstantHacker. You’ll find links there to what we’re doing in Quantum Computing Services at Protiviti. You can also DM me questions or suggestions for what you’d like to hear on the show. For more information on our quantum services, check out Protiviti.com, or follow ProtivitiTech on Twitter and LinkedIn. Until next time, be kind, and stay quantum-curious.