Summary
Mark Jackson, senior quantum evangelist at Quantinuum, is interviewed by Yuval Boger. Mark and Yuval talk about optimizing compilers, quantum error correction news, whether customers prefer shrink-wrapped quantum software, and much more.
Key takeaways
- A well optimized compiler can cut gate counts and errors enough to rival gains from better hardware, so software stack quality is not a side issue to qubit count, it competes with it directly
- Customers evaluating quantum computing mostly want packaged, shrink-wrapped software rather than raw low level access to qubits and gates, which changes how vendors like Quantinuum prioritize their roadmap
- Recent quantum error correction results are treated as a genuine milestone in the conversation, not just incremental progress, and are discussed as a signal of where the field's timeline is heading
- Quantinuum's evangelist role involves translating between what the hardware and compiler stack can actually do today and what potential customers assume or hope it can do
Read the full transcript on The Quantum Computing Report
Transcript
Yuval Boger: Hello, Mark, and thanks for joining me today.
Mark Jackson: Hi, Yuval. It's a pleasure. Thanks for having me.
Yuval: So who are you and what do you do?
Mark: So I am the senior quantum evangelist at Quantinuum, and Quantinuum is the recent merger between what was Cambridge Quantum, which developed quantum software, and Honeywell Quantum Solutions, which developed quantum hardware.
And so as an evangelist at the company, my role is to do a lot of the public-facing activities, getting people excited about quantum computing and getting them to understand how they can benefit from it.
Yuval: And I know you come from the CQC, the Cambridge Quantum Computing side.
Mark: That's correct. Yeah, I actually just had my five-year anniversary at the company.
Yuval: Congratulations. Tell me a little bit about the CQC products, if you could.
Mark: Sure. So CQC — although of course we're one company, we are Quantinuum — we develop products in every area of quantum software. So this includes a software development kit, which is now open source, and we call this TKET.
There's also an open-source tool called lambeq, which is for computational linguistics and natural language processing. And then on the commercial-facing side, we have chemistry, and this includes a computational chemistry package that we call InQuanto. There's also machine learning, optimization, and then finally cybersecurity.
And so in pretty much every area of these different applications in quantum, we have a team of experts.
Yuval: When you interact with a company, say a pharmaceutical company — not a company whose main business is quantum, but rather one that wants to take advantage of quantum computing — do you feel that they are looking to write their own algorithms, or do they just want a sort of shrink-wrapped software package that they could plug a molecule into and get the information they need?
Mark: That's a really good question, and I especially like that you used pharmaceuticals as an example here, because personally, that's actually the application that I'm most excited about: how quantum will affect medicine. Right now it takes about 10 years and a billion dollars for a pharmaceutical company to bring a drug to market, and that's because they can only use a computer in the very preliminary stages to try to guess generally how a drug might behave with our biology.
But after that, they have to just do clinical trials. They have to try a lot of potential drugs, try them on people, see what the effects are, and hope that there's a generally good effect and that the side effects aren't too bad for the general population. And so of course this takes a lot of time and money, and it's expensive.
And it's very risky for the people involved, of course. And so ideally a pharmaceutical company could use quantum to quickly hone in on the most promising medicines, and then only at the very final stages would they have to do the clinical trials, after a lot of the preliminary work has been done.
In regard to your question — do I think that they would want to be quantum experts themselves, or would they want to just kind of plug and play? I think both, actually. I think there are some larger pharmaceutical companies who would want to be experts, and they would have in-house teams to do this type of thing.
And there are others that would be fine with just using quantum software as a service. And we're already starting to see some of these types of models emerge. So yeah, I think both are important.
Yuval: I had the pleasure of interviewing a couple of pharma-related quantum companies, and they talk about quantum advantage and being able to do things that you couldn't do classically. You mentioned 10 years; if that could be shortened with quantum, that's very valuable to humanity and very valuable to these companies.
How soon do you think that turns into reality?
Mark: There are indications that we will see commercial applications in as soon as three to five years. Of course, people have different opinions, and that's not a guarantee, but there is reason to think this based on the current trajectory of progress. The metric that's being used right now for the power of a quantum computer is called quantum volume, and it's a complicated formula.
I can't easily express it, but it's a rough estimate of the power of a quantum computer, and our company, Quantinuum, has increased the quantum volume by a factor of 10 every year for the past several years, and we think that we can continue with that for the next several years. And so based on that, we think that in the next three to five years, we will start to see commercial applications.
If you were to ask me what the first commercial applications would be, I would think either chemistry — so pharmaceuticals, as we just talked about, would be an example — because in chemistry, you see this exponential speedup compared to classical processes. And the other application coming soon, I think, would be financial machine learning, and the reason for this is because in finance, when you have these trading algorithms, you don't need to be correct 100% of the time.
Even if you have just a small increase in accuracy, that's very valuable. And so if we could use quantum machine learning — and in particular the quantum Monte Carlo algorithm, where our research team actually showed that there is the quadratic advantage people had believed there should be — then we should start to see advantages in the not-too-distant future.
Yuval: You mentioned lambeq and now Monte Carlo simulation.
If I understand correctly, these software packages are written on top of TKET, your optimizing compiler. Could you tell me a little bit about TKET?
Mark: Certainly, yes. All of our software is built to take advantage of TKET. What TKET is, is a quantum software development kit which allows you to take your quantum circuit, your quantum program, and run it most efficiently on any platform, either hardware or simulator.
And so there are two components to this. The first component is just taking your program, the circuit, and looking for ways to simplify it while keeping it mathematically identical. A quantum program is really just a very sophisticated set of operations manipulating the qubits. And so TKET looks for patterns where it can get exactly the same end result, but in a simpler way.
It would use fewer operations to achieve that result. It has a very sophisticated set of optimization routines that it goes through, and once it does that on the circuit, it then finds the most efficient way to execute it on the hardware platform that you've chosen. And the reason that it's different on each platform is, first, that qubits can be connected differently.
So on the superconducting platforms, for example, the qubits are physically locked in, and they only connect to their neighboring qubits. The second thing is that the gates which are native to each platform are different. Every platform has the operations which are kind of built into it, and these aren't necessarily the operations that you would use in your circuit, of course.
So some sort of translation often needs to be done. And third, there are advanced features which are starting to pop up, like mid-circuit measurement, which some platforms, including ours, can do. And TKET is aware of this. It has a backend file for each platform, and based on that information, it will most efficiently execute your circuit on that platform. It will do all of this translation and the routing — moving the information between the qubits — all for you.
You don't need to do anything by hand. We have developed TKET since the beginning, about eight years ago, and it was so valuable for us in our software development that about a year ago we made it open source. So not only is it free, and anyone can use it however they wish, but every line of code is now available.
So you can inspect exactly how we do things. If you have some clever idea for how to improve it, then by all means, please get in touch and submit your idea. If you have a hardware platform, then you can write your own backend extension so that people can run software on your platform.
So yeah, that's a quick summary of what TKET is.
Yuval: You mentioned connectivity for TKET. How about noise? Different computers might have different noise models based on different technologies. Does TKET look at noise and optimize to minimize it, or to get the best fidelity for the circuit?
Mark: Absolutely, and I'm glad you mentioned that. So yes, in the backend extension file, you can input the characteristics of different qubits on the platform. And TKET will automatically try to use the best qubits and avoid the worst ones, increasing the performance. So yes, that's exactly right.
Yuval: And is that true just for qubits, or maybe some gates are noisier than others, or maybe there are other optimizations you can do?
Mark: Yes. You can finesse all of these different optimization parameters as you wish.
Yuval: How does it feel to work in a company that does both hardware and software? Is that a plus for customers, or is that a negative?
Mark: It's definitely a plus. In fact, we are the largest quantum computing company in the world now — pure quantum, that is — and a big advantage is that we can do everything in-house.
So when a company approaches us and says they want to work on a project with us, we have the scientific expertise to design the algorithm, write the application, and then test it using our own hardware. So this isn't some extra thing that needs to be negotiated, or an extra fee tacked on, or anything.
We can do everything in-house. And then of course, scientifically, it's great that we have the hardware experts and the software experts all working on the same team, and they can communicate.
Yuval: You've been in the quantum industry for at least five years, I believe, and you've seen it grow. What do you know today about the quantum industry that you didn't know or didn't realize six months ago or 12 months ago?
Mark: It's going faster than I thought, to be honest. One example of this is that just a few weeks ago, Quantinuum announced a big breakthrough relating to quantum error correction. And just to quickly give the context for this: qubits are physical objects. There are different ways to build them, but they're physical objects, and so they don't always give the correct answer.
There are a lot of sources of error. And if an error happens, then of course that kind of spoils the result of the circuit, right? That gives you the wrong answer, and so you want to avoid this. Now, this happens with normal computers too, of course. Transistors are physical objects, and so once in a while they might give the wrong answer.
But with normal computers, error correction is a very simple process. You simply triplicate every bit. So instead of one bit, you have three. In the unlikely event that one bit fails, you just take the majority. You take two out of three, and you disregard the third one if it's doing something different.
Now, the reason you can't do that so simplistically with a quantum computer is that quantum physics has all these funny rules about no copying of quantum data, and you can't even measure a qubit without changing its state, of course. And so quantum error correction is a very subtle process.
Fortunately, someone did figure out how to do this quite some time ago. It was Peter Shor, the same Shor as in Shor's algorithm, and he gave the first example of quantum error correction. We now know of several ways to do it, but the basic idea is that you smear the information of one qubit over several qubits, and then you very cleverly, indirectly test whether an error has popped up, and if so, you correct it.
And so we've known mathematically how to do this for some time. The trouble is that this quantum error correction process — the smearing process and then the correcting process — takes operations, of course, and you're introducing more qubits into the scene. And so if there's a relatively high chance that a qubit will fail, you've actually made the process worse.
It's actually more likely something will go wrong, because you have all these operations you're doing now, and there are more qubits. There's more potential for something to go wrong. And so you actually have to start with a pretty low error rate to begin with for this quantum error correction process to help.
Otherwise, it actually hurts you. So getting back to what I mentioned, just a few weeks ago we announced the first breakthrough point where we were able to do quantum error correction in a way that helped and didn't hurt. And this is a huge milestone. Obviously, a lot of work has to be done.
We want to keep going so that it really helps us, but this is a very significant milestone for the industry in general. If no one had ever reached this milestone, there wouldn't be any hope of having practical applications. And so it's really amazing to see this kind of historic achievement done just recently.
Yuval: So you're basically saying progress is faster than you imagined, and as a result, perhaps the time to quantum advantage, or at least to true quantum usefulness, has become shorter.
Mark: Absolutely, yes. And I think the industry is seeing that. If you look at all of these metrics — the investment that's going into it, the number of quantum companies that are being created now, the number of projects that are being done — all of these show this hockey-stick type of behavior, where it was very slow for a long time, and then just in the past few years it's exploded.
And when I talk to companies, and they ask about quantum and are kind of curious about it, and I tell them, "Well, I think we're only a few years away from having commercial applications," sometimes they respond, "Okay. Well, that's good. I'll come back and talk to you in a few years." And the problem with that is that it takes time to do these things.
It takes time to develop software, and it takes time to build up this quantum muscle, if you will. You have to hire the teams and get them to understand not just quantum computing, but how it relates to your particular industry. And you can't do this overnight. It takes a few years.
And so the companies that are going to benefit most from quantum computing have already started. If whoever is listening to this is in a position to get into this, it's not too late, of course. I would encourage them to get into it. But the idea that you can just flip a switch overnight and suddenly become quantum ready is not true.
It does take time to do this.
Yuval: As an evangelist, how do you walk the line between exposing what the future could look like and avoiding hype? On one hand, you see articles saying quantum computing is going to solve all the world's problems, and on the other, "Oh, it's a bubble, it's hype, it's just going to crash and burn."
How do you walk that line?
Mark: That's a very good question. In fact, I'd say that about 99% of the people I speak with outside of our industry have a misunderstanding about quantum, and they're opposite misunderstandings. It's actually the two that you just mentioned. About half the people I speak with don't think quantum actually exists.
They sort of think it's like science fiction and we've kind of made this up or something. And then the other half think, "Oh yeah, quantum is magic. It means that you just take your existing software, you run it on a quantum computer, it's a million times faster, and we can do that tomorrow."
And neither of those is true. It's good that I have a physics background. My background is actually in superstring theory and cosmology; for 15 years, this is what I researched. So I have a bit of technical background about what quantum is, and the reality is that quantum computing absolutely does exist.
There are many quantum computers in the world. We have a lot of quantum software that's been developed now. There are many companies, many startups, and many academics developing quantum software. We can do interesting things already. Nothing is life-altering yet, but as I said, I think we will start to see commercial advantages and applications in just a few years.
But quantum is only for very specific applications. By no means is it a general speedup where you just take your existing software and run it on a quantum computer. The logic behind quantum software is completely different. Quantum physics works in a very non-intuitive way, so you have to start from scratch to develop these algorithms.
We have no general formula for how to write quantum algorithms. We know of a few quantum algorithms which show promise, but there's no general formula for developing them, and we're still figuring out what the applications might be. It's entirely possible that next week some very smart young person thinks of some killer app that no one has conceived of yet.
So yes, what you asked is a very good question. A large part of my role as an evangelist is to get people excited about quantum computing, but also to manage their expectations about what it is and what it's not. That's a large part of my job.
Yuval: As we get closer to the end of our conversation today, I'm wondering: if you had a magic wand and could control what the industry is doing, what would you like the industry to be doing that it's not doing today?
Mark: Actually, nothing comes to me off the top of my head. But I will tell you what I am proud of that the industry is already doing, and that's the big emphasis on open-source software development. I already talked about our TKET software development kit. I mentioned lambeq earlier, which is a computational linguistics and natural language processing toolkit that we've made open source.
There are a lot of other tools, like IBM's Qiskit. There's a nonprofit called the Unitary Fund, which is dedicated to open-source software development. And so I'm very happy to see people from a lot of different groups, both commercial and academic, coming together from different angles and working on this.
Yuval: Mark, how can people get in touch with you to learn more about your work?
Mark: My email is mark.jackson@quantinuum.com, but I'm also available on LinkedIn, of course, and through our company website.
Yuval: Thank you so much for joining me today.
Mark: Absolutely. Thank you, Yuval. It's always a pleasure chatting with you.
