Podcast with Dr. Robert Sutor, IBM

My guest today is Dr. Robert Sutor, Chief Quantum Exponent for IBM. Bob and I discuss why IBM is in the quantum computing market, what IT managers should expect when integrating quantum into the enterprise, shares his market predictions and much more.

‍

Listen to additional podcasts here

THE FULL TRANSCRIPT IS BELOW

Yuval Boger (Classiq): Hello, Dr. Bob, and thanks for joining me today.

Dr. Bob Sutor (IBM): Happy to be here on a cold, October rainy morning in New York.

Yuval: So who are you and what do you do?

Bob: Who am I? Well, I've taken a long time to try to figure that one out. I'm Bob Sutor, I'm a member of the IBM quantum leadership team. Long-term IBM person. I'm a mathematician by training. Very recently I came back to research nine years ago, IBM research, to lead the mathematical sciences department. And then I noticed way down at the other end of the building, a research headquarters in Yorktown Heights, New York, there was suddenly all this activity. And that's where all the physical sciences people, the physicists, the engineers ... I was a math and computer science guy, right.

But there was a lot of commotion down there. And as I learned more, it was the IBM quantum program really being born in the sense of having a quantum computer. So I switched over to that a few years ago, was kind of a vice-president at large, I would say, with the business aspects as well as the technical.

And then recently I've been, I have this title, Chief Quantum Exponent, which is a little bit of a play on words. I spend a lot of time talking to people and writing about quantum computing. As you know it's a nontrivial concept, and coming from a very interesting part of science, quantum mechanics. So in some ways I translate that science and the computing aspects to tell people about what quantum computing is going to be good for once we get there.

Yuval: Got it. So, IBM doesn't make as many computers as it used to many years ago. Why is IBM in the quantum computing market to begin with?

Bob: Well, I think, if I may qualify your statement a little bit, if we look at pure quantity, yeah, we used to make IBM PCs. I mean, we were pivotal in the early eighties in really launching the business personal computer market. But if you look around at the financial institutions around the world, I mean, these are still driven by IBM Z computers. Until very recently, the largest supercomputer in the world was driven by IBM power technology, right? Along with NVIDIA GPUs and things like this.

So don't be fooled to think that IBM is not a very major player in the IT industry. And of course, in the cloud industry as well.

Well, fundamentally we're a big company. We've been around for over a hundred years. And we do computing and what it's good for. I mean, that's what it boils down to. Since we have been around for a while, we haven't just been thinking about this quantum idea for the last few years. In fact, our early research goes back to the 1960s. Charlie Bennett, who's still an IBM fellow, coined the term quantum information science in February, 1970.

So arguably IBM has been in the quantum industry for over 50 years. We were the first ones to put quantum computers on the cloud in 2016. We now have 25 quantum computing systems in the cloud.

We've retired more than 25 systems. We have retired more systems, older systems, than probably everyone else has all together, right? So we are in it because we think quantum computing is a very important element to the future of computing. And I personally would go so far as to say, it's likely to be the most important computing technology for this century.

Yuval: So if you look at how quantum computers are expected to be deployed in the enterprise, I think it's unlikely that it'll be just a pure quantum play, right? I mean, we don't think we're going to, in five years or 10 years, run a Zoom call on a quantum computer. It has to be somehow meshed within other types of IT infrastructure.

Do you expect the IT department to just manage the quantum relationship just like they manage the cloud relationship? Do you think there are any special requirements for quantum to integrate it into the cloud or the data center?

Bob: Well early on, and here I mean four years ago, people kept using this word hybrid when they talked about quantum and then the traditional systems. Which in this industry, we usually call the classical systems, right? So those are the, those chips, classical chips are the ones that are in your phone, your laptop, in the super computers and things like this.

They kept using this word hybrid. And I never liked that. I always felt uncomfortable with this term. And in fact, there was this confusion about, will quantum computers replace these classical computers? And there's really no need. I mean as you pointed out, whether it's Zoom, or the interface for your phone or something like this, classical computers, and probably what's more important, the classical computing model, think of the programming model, works very, very well for many, many things.

So instead we are talking about the integration of the technologies. Use each for what they are best at. And in fact, that typically will mean that classical algorithms will drive the use of quantum algorithms in that light. Now to the degree that your average IT person is aware of everything that's going on in their data center, every time they reach out to other clouds, every time they hit the server, no.

At some point quantum will just blend in. But it's this extraordinary power that we anticipate quantum systems to have for certain types of problems, that's where they will be aware they're being used. We won't use quantum computers for problems for computations that classical computers do very well right now.

Yuval: So you expect the IT managers to just over time learn to use, learn how quantum computers work? Or how the vendors provide them. And it's just going to be part of the IT fabric. Is that correct?

Bob: Well, when we talk about it, you have to separate hardware and software. So when we talk about software, and again, just to pick a common example of your phone, a smartphone, whatever brand it is. And I happen to have an iPhone. So if I'm writing an iPhone app, and I want to draw a line from one corner of the screen to another, there is a very high level routine that I would call. I would give it the coordinates of the beginning and the end. I would say this is the color of the line, this is the width of the line, and away it goes. And the line appears.

I don't worry about the very, very low possible level way of placing all of those dots on the screen. Because there is a hierarchy of APIs, of functions that get easier and easier to use. So in the same way, the people that we call the model developers, the people who will be calling these higher-level functions that use quantum, don't necessarily, or won't necessarily have to know the lowest-level bits of how you talk to the quantum hardware.

So from an IT manager perspective, no, they're going to be working at the higher level. But if you're talking about people who are building the highly-optimized routines and algorithms that others will use, yes, they will worry about the hardware.

Yuval: If you look at the cloud providers today, whether it's Amazon, or IBM, or Google, and so on, on one hand, Amazon would sell capacity: “oh, you need another EC2 instance, here you go”. And on the other, there's an API interface. It could be a maps interface for Google, it could be a speech recognition interface. What do you envision the primary usage method for quantum would be? Would it be, here's a computer, submit the circuit and it's going to run? Or for instance, here's an optimization API, give me your TSP graph, and here's the solution.

Bob: I believe it will shift over time. So in the same way that we just discuss the stack and low level programming. At the beginning, when new computing processors come out, there's a lot that you work at, at a low level. So for example, from a programming language perspective, C is pretty low level. C++ higher level, Python is higher altogether. I don't worry, especially about the details of the hardware when I'm using Python, but I'm much more conscious of it when I use C, let's say.

So for most people, eventually they will be using the higher-level routines, called this function. And there'll be relatively few people who work at the lowest level. In the same way that there are relatively few people today who work at the lowest, what we'll call assembly language level, for your phones or your laptops, right? There are some.

But at the beginning it was a completely reversed. It was, to put in the terms of, as you said with quantum, people would develop circuits, right? People would call the circuit compiler on it. They would go to the computer, to the quantum system. It would come back, and they would understand about that.

But we've already been building algorithms for financial services, for chemistry, and so forth. These will translate over time, as we've said in our roadmap. Well, IBM published a, what we call a development roadmap February. These will become part and parcel of quantum cloud services. Where yes, you call a high level API saying, here's my data, right? This is what I'm trying to accomplish. Go off and do it. Now, there's a lot of smarts in how we've programmed the, go off and do it. But that's how most people eventually will use quantum computers.

Yuval: We see an increasing number of enterprises start exploring quantum, and I'm guessing you see the same. What do you see as the biggest barrier to accelerating the adoption? Is it just more qubits with less noise? Is it a software framework? Is it the people angle? What do you see as the biggest issue that's prohibiting faster adoption of quantum computing?

Bob: It's awareness and understanding what is and is not true about quantum computers. Because there is a lot of hype out there coming from some corners in the market. It is, as part of that education, to learn a little bit more. And then there's hands-on skills development. Because for those who are completely waiting, saying, well, there'll be ready eventually, that's when I'll start. Well, I've got news for you, seven of your competitors have already started. So by all means, wait, if that's your strategy, right?

In terms of the use of the systems to do better than what classical systems can do, it's really a question of three things. So first of all, you need enough qubits. So a qubit is the basic information unit, it's implemented in different ways by different vendors. IBM uses a superconducting transmon approach to this, which is the only approach so far which seems to be able to scale out the very low double digits, right?

So we need enough qubits to be big enough for whatever problem you're trying to solve, right? Which means you don't need two qubits, you don't need five qubits, you don't need 11 qubits. Ultimately, you need hundreds of thousands of qubits like this. They need to be very good qubits. They need to be very high quality qubits. And this is just an artifact of how these systems run. They're based on quantum mechanics, but so is the rest of the universe. And the rest of the universe is really trying to mess up your computation. So we need a certain amount of isolation, noise reduction there.

We need speed. So it's one thing to have qubits, but if one type of qubit is hundreds of times slower than another type of qubits, you're really coming into the advantage here. So it's quantity, it's quality, and it's speed. And once we get the right balance of the three of these, we can start using today's algorithms, so called noisy algorithms, to start getting some interesting results, and ultimately get full tolerance and error correction. When we can start implementing all those algorithms you read in all those quantum computing textbooks.

Yuval: So, IBM has published a roadmap of what you expect to see in quantum. But if I gave you a magic wand and you could control software companies, or hardware companies, or other players in the industry, what would you have us focus on for the next 18 months or so?

Bob: I think for the major work that people need to do is, as I said before, it's increased awareness. It's increased education. It's increased skills. And I will add a fourth, which is experiment. And by experiment I mean, try to map your use cases to an integrated quantum classical approach. Because I'm sorry to say, no one is going to walk in today and say, here is a pure quantum approach, or a complete quantum approach, that does your problem far better than everyone else. In spite of what some marketing material may claim, right? So it's just not the case.

So you've got to experiment. As the system scales in the way I talked about in terms of quantity, number of qubits, the quality and the speed of qubits, you have to get on the right track to understand, is this a good approach that will eventually handle my use case?

And I would even go a step lower here, because we toss around this term use case. And so for example, risk assessment, or risk analysis, or something like that. What does that mean for you? I mean, that's a very high-level term, right? Where does this idea of identifying risk fit into your actual workflow? So we have spoken for a long time about these high-level use cases; the chemistry, financial services, AI. We've got to go deeper. So all these other companies, yes, increase education, skills building. But help clients experiment and get on the right track, and understand very specific parts of their workflow where quantum will eventually plug in and make a difference.

Yuval: We're at the last quarter of the year, and so this is prediction time. What do you predict to happen in 2022 and 2023, as it relates to quantum computing?

Bob: Okay, well, I want to put in a little plug for something IBM produced this last year. And it is a book, it's available in PDF form. And it's called The Quantum Decade. And if you just search for that, you can download it. Do bear in mind that it is a printed book, it's a beautiful 12 inch by 12 inch book. So as a PDF it's 120 pages long. But it's a beautiful PDF, and it talks about what we think will be happening, and what people should be doing in this decade. Because we do believe that the 2020s are going to be the decade that are really significant for getting quantum off the ground.

But more to your question. So for next year and the year after, I'm going to go right back to our roadmap. So the end of this year we will release a quantum computer that has more than 100 qubits. Next year, over 400. And in 2023, over 1000.

And so what does this represent, first of all? Well, it represents our confidence in our science and our engineering that we have broken through the scalability roadblock that happens with qubits, and quality, and so forth in things like this. For your listeners who may not be that familiar with quantum, adding more qubits is not just a question of manufacturing, right? If I buy a laptop and it has eight gigabytes of RAM, I order online another eight gigabytes of RAM, and I plug it in, and away I go.

Well, it's not quite so easy, right? qubits all have to be able to talk to each other. So there's a complexity involved with doing this, it's not simply additive. So it becomes harder and harder. So we will break through the hundred-qubit boundary this year, 400 next year, and a thousand.

This means that we can start to make a serious run at what we call quantum advantage, which at the beginning will be particular cases where we can demonstrate that these integrated quantum classical systems can do better than classical will. I'm not guaranteeing it will happen, I'm saying people can make a run at it because the systems are starting to get big enough to do this.

I also think at that point there's going to be a little bit of, well, I was going to use the word shake out, and I don't quite mean that. But I think there's going to be a better understanding of which qubit technologies really are going to be the most promising. Once you start getting some things where, again, very low double digits, and they seem to be stuck there, and you start getting other technologies, which goes into the hundreds and thousands it's, people are going to make their choices.

Yuval: So Bob, how can people get in touch with you to learn more about your work?

Bob: Well, I'm on LinkedIn, that's the best way. Robert Sutor. I publish pretty frequently about quantum, various things like that. If I may do a personal plug here for a moment, I have two books out over the last three years. The first is Dancing with Qubits, which is, for the most part, a non-physics approach to learning about quantum computing and the algorithms. I bring you along, I show you the math you need. And by the end you'll understand that at least enough to get started with quantum computing.

And then a little bit more than a month in the market is Dancing with Python, which is an introduction to coding, but where I teach you classical coding and quantum coding at the same time. Because at some point we have to stop thinking of quantum computing as this add-on. It's computing. And I'm saying this, yes, it's a plug. But it's also a warning. Is that by all means, communicate, connect with me on LinkedIn. But I may be talking about this latest book Dancing With Python a bit in my postings.

Yuval: So Bob, thank you so much for dancing with me today. And thanks for joining the podcast.

Bob: My pleasure. It's been a great conversation, thanks.