Superposition Guy’s Podcast — Bob Sutor, the Futurum Group

Xpanse Xpanse

“The Superposition Guy’s Podcast”, hosted by Yuval Boger, Chief Commercial Officer at QuEra Computing

Dr. Bob Sutor, a 40-year industry veteran with a rich background in IBM and other tech companies, discusses his current role as an analyst and advisor at the Futurum Group. Bob reflects on his diverse career from mathematical sciences to quantum computing, emphasizing his shift from a vendor to a broader industry-wide perspective. He also mentions his recent book update, “Dancing with Qubits,” aimed at demystifying quantum computing for non-physicists, and shares insights on the current state and future directions of quantum computing.

Listen on Spotify — here

Full Transcript

Yuval Boger: Dr. Bob Sutor, thank you so much for joining me today.

Responsive Image

Bob Sutor: Glad to be here. Third, third time’s the charm, right?

Yuval:  Absolutely. You’re the podcast champion for me. So who are you and what do you do?

Bob: Who am I? Yeah, well, you know, we just mentioned three times. I’ve worked for three different companies for each time over the last few years. However, I will say that the first one, IBM, 39 years as far as IBM Quantum. I am a 40-plus year industry veteran. As I just said, I spent most of my career at IBM. About a third of it on the business side, again, software, but most of it in IBM Research.

And I was the head of mathematical sciences at IBM Research and then part of IBM Quantum. IBM, of course, is a very large company. I thought I’d try something a little bit smaller. So I went to ColdQuanta then renamed Infleqtion for a couple of years, about 200 people. And for the last four and a half months, I’ve been working as an analyst and an advisor at the Futurum Group. So I now have this very different perspective that I am not a vendor. So people like you Yuval, you know, who are vendors, you just tell me all your secrets left and right. It’s wonderful.

But it is really nice having an industry-wide perspective, what’s happening with quantum and also cover AI. I try to cover the more advanced areas for artificial intelligence.

Yuval: I do tell you, but not all my secrets. I don’t want to get fired quite yet.

Bob: I’ll take that part back then, but not really.

Yuval: And then I also know that you’re a book author and there’s a second edition of something coming out or that came out recently, right?

Bob: Yeah, it came out a few months ago. A long, long time ago, I co-wrote a book like literally 30 years ago. And when I was learning quantum computing around 2017, well, there were a few books out. And I said, you know, I’m a mathematician, I’m not a physicist. There seemed to be this real assumption that you had to be a physicist in order to understand quantum computing.

So I didn’t really find a book that I liked, so I wrote one. I wrote a book that I thought would appeal to people like me coming from a mathematical perspective. And that was called Dancing with Qubits, the first edition. I followed that up with one called Dancing with Python, which is an introduction to programming using Python, both classically and quantum. But then it came time to do the second edition. And as I say, the second and final edition of Dancing with Qubits, because about four or five years had gone by and some things I thought would change didn’t. Some things that I thought, you know, I thought NISQ would be further along. Let’s put it that way. Or closer to its ultimate demise in favor of error correction. It wasn’t. Quantum machine learning – there’s a lot more interest in it. I do think it’s primarily research though at this point. Just to be clear, I don’t think it’s close to being commercial. And then I taught a semester with the book. And so whenever you actually teach with something, you realize, well, okay, maybe I should put this section in that chapter and reverse this. And I really don’t talk about that very much. So I think it’s a much stronger book, Dancing with Qubits, second edition from Packt publishing. And now, I’m going to take a break now for a long, long time on writing such technical books.

Yuval: Very good. So what is the state of quantum now that you’re five months, of course, all your previous experience and now five months at the Futurum Group, what do you think is the state of quantum these days?

Bob: I see quantum, you know, really in three phases. And by the way, lots of people like to have three of them. But in this case, I do think it’s reasonable. I think the first phase of quantum is, hey, let’s just try lots of things. Let’s give a lot of people money to try all sorts of variations. Which is okay. But at some point you have enough learning and you say, all right, well, this flavor of that modality is probably going away. And it’s scaling, but maybe not in the way we think, right? The fidelities aren’t what we expected. The control is a little harder. I tend to think every modality has one very serious problem to it. So it could be coherence time. It could be the number of qubits that you can have in it. And so forth and so on.

The second phase is more like, okay, we’re much smarter now, but what is the intention here? The intention is to have a very large and powerful quantum computer that can do some of these things we’ve been promising for years. And just to fix ideas, let’s say a hundred thousand really good physical qubits. There are all sorts of theoretical arguments of how many qubits we need to do this, that, or whatever, or Shor’s and things like this. Let me just pick a hundred thousand as a reasonable number. So what do we have? Well, IBM did about 1,100, a little bit more than 1,100 last year experimentally, but they’re focusing more on between one and two hundred. Just about everybody else is less than a hundred, different degrees, but they’re all examining ways to get more. And different modalities will be able to have more, they believe. Though it’s much harder to control 500 of something than it is five or 50, right? So there’s a fair amount of hand-waving.

But what we have to do in the second phase is really to be very practical and say, we have to make a few hard choices. We have to start thinking of all the auxiliary things. Are we going to need cooling? I think almost everybody’s going to need some cooling somehow, some way, whatever room temperature means. I like to say, my refrigerator in my kitchen is room temperature, not on the inside, but on the outside. So some cooling is involved with this, but we have to get practical and say, we need this technology. We have to invest here. If we don’t get that technology, for example, of connecting multiple quantum processing units, we’re going to die because they’re all going to be too small. The world will be littered with tiny little quantum computers. So the investment has to start shifting away from a few dozen qubits to solid, well-performing QPUs that are connected together. And then on the software side, there’s a lot of compiler technology, how do you optimize across these types of things? So I think the next few years are going to be very interesting.

I think there are a few companies doing really exciting work in this area, and I hope they continue to do so. I think we’re going to see the rise of open source in a lot more projects, the software side of the projects in ways we haven’t so far. I think there are some interesting recent changes. IBM with Qiskit 1.0, there was a lot of rewriting done with that, a lot of redoing the infrastructure with Rust for performance. Because if we are moving to a world where we have to code hundreds to thousands or tens of thousands of qubits, your environment has to change, your libraries have to change. You have to be much, much more efficient. So we are still in the area of a lot of experimentation, but we’re moving into the area of let’s buckle down and invest in the essential technologies, which doesn’t mean yet another few qubits made in some sort of interesting technology.

Yuval: So is that the second or third phase?

Bob: That’s the second phase. The third phase is, yay, we win. So that is, we’re not just doing laboratory quantum supremacy or advantage or things like this. We actually start to see quantum computing being used in production, at least somehow in some way for some interesting problems we care about. That’s the third phase. That’s not all going to arrive in one day, just like error correction doesn’t all arrive in one day. But the second phase is the, I’ll call it as I just said, the buckle down, let’s build the essential technologies. The third phase is we put them together and actually start seeing some of these advantages for interesting problems.

Yuval: You mentioned that you had some surprise between the first edition of your book and the second one about the progress of NISQ and how close it would be to its demise. What else has materially surprised you over these last couple of years?

Bob: So NISQ is, just to be clear, I’m sure people listening to this podcast know this, but just for clarity, it’s the noisy intermediate-scale quantum. The idea is you’re primarily dealing with physical qubits that do not behave perfectly. They’re only valid for a certain amount of time. Then they become chaotic. They lose their values. And so you have to rush and do a few things before errors really completely destroy the values. The next big thing after NISQ is error correction, fault tolerance and so forth.

So what I’m surprised about is across the board, why don’t people have more qubits now? So I started that last book in 2018. It was published in 2019. Lots and lots of people still have 20, 30, 40, 50 qubits. Years later, they’re getting better, the qubits are getting better, but we’re not seeing hundreds and hundreds of qubits. IBM is the only one that really went that high. There are some reports from some other countries, but so that’s one thing. I don’t think the NISQ algorithms themselves are particularly working out and it’s because of the problems of coherence time and errors and things like this. They’re repetitive iterative algorithms. If you can’t iterate, you can’t optimize. It’s almost like a neural network that you can’t use back propagation to actually optimize and things like that. So people thought that the NISQ algorithms would work out. We probably see some interesting things. I don’t think they are for the most part. And for that reason, I don’t think they are for quantum machine learning.

So what I did do in this book was specifically add two chapters, one on NISQ and quantum machine learning. I do survey some of the algorithms within the tone and the level of the book, but it was cautionary. It’s like, I’m not telling you this because it solves all your problems. I’m telling you this is just, this is what people are trying to do, but this is why it’s not working out. This is why we do need some of the other things like that. I had been hoping that NISQ and AI would go away before I wrote the second edition. And let me just say from an exposition perspective, most people in the world don’t think about Hamiltonians all day long. I think sometimes physicists think they do constantly. I don’t. So whenever you have to re-express something in terms of Hamiltonians and Hermitian matrices and things, you’re going to start losing people. So this was one of my challenges in writing the book, to say enough about this.

Yuval: You spoke about many companies investing in quantum and on the other hand that you’re hoping they’d be farther along, whether the number of qubits or error correction. Do you think it’s time for the government to step in? Do you think there needs to be some kind of an Apollo program for quantum computing?

Bob: So I do. And let me just say I will be leading a session at the Quantum World Congress in DC, September 9th. And we’re going to bring people together to talk about that. The Apollo program really was three programs, Mercury, one person. Hey, here’s a missile. Let’s see if we can send it up with a person and bring them back down. Gemini, what I call the let’s buckle down and then Apollo, let’s go to the moon and come back. So I do think we need that. And so really for today, I’m just planting the seed of that and I hope you’ll join us, those of us, and I’ll be talking much more about that.

I generally think there’s a tension between government funding and private funding with quantum. Government thinks private investors should put in the money and the private investors say, hey, this is going to take really too long beyond our ROI horizon. So for the beginning phases, the government should be investing. I believe we’re just going to have to start, they’re going to have to come to some sort of agreement. I think there are too many companies operating too independently. We see announcements all the time of partnerships. I don’t put much credence in announcements of partnerships. I love to see the results of partnerships.

So I think under somebody’s direction, and it could be private, but saying you five companies who have complementary technologies are going to work together and have a fully functional unit that does all of this. And maybe they do that by investment and control this way. Maybe there’s a little bit of government. Government often controls things via their purchasing power. So if they can guarantee that they will purchase so many things at a certain price, well, maybe the money to actually develop it will come from elsewhere. But we need something. Otherwise, at least in the United States, I think we’re going to continue to have hundreds of seemingly random disconnected quantum companies.

Yuval: In your current job, you also cover AI. How long before quantum becomes at least a part of the AI workflow? Let’s not throw out all the GPUs quite yet. You have a quantum processor that does something useful in an AI workflow.

Bob: So if I were the CEO of a quantum AI startup, I’d say two years, no problem. But if I was more pragmatic, I would say it’s a whole lot longer than that. For this reason. Quantum does not handle very much data. It can’t. The machines are small. And even with the exponential factor of doubling the amount of information you can have with every qubit, it could be as bad as exponential time in loading classical data into a quantum computer. So you have this classical data. You need to put it into a quantum computer to do something interesting. But you run out of time just loading the data. You don’t have time to do anything with it. So that’s a little bit of a problem.

Like gee, I put 5% of my data into the quantum computer, but then it couldn’t do anything. Well that’s not really a solution. If you happen to have the data that’s quantum encoded already, such as saved from a computation, or maybe coming from a quantum sensor of some sort, you have time series quantum encoded data, well, maybe that’s becoming a little bit more interesting. The second thing is, as we talked about before, many of the quantum machine learning algorithms are variational. In effect, once again, the coherence time bites you and you ultimately can’t do that much better. So therefore you’re limited to small problems. And you know, quantum’s not for small problems. There’s a threshold. Classical does just fine for problems, often very big problems.

In the AI world, for example, I was at Dell Tech World a couple months ago, and they were talking about a generative AI data center that took up six acres of space with 32,000 GPUs. All right, so dedicated power, dedicated compute, dedicated cooling. That’s an awful lot of processing power. So this notion of a supercomputer is getting diluted because of all these great big Gen AI data centers that are being built with many, many GPUs, which means that the problems that you must have quantum for are getting bigger and bigger and bigger and bigger, in particular, quantum for AI. So of all the use cases for quantum, such as modeling physical phenomenon, chemistry, quantum chemistry, maybe some of the optimization work, but certainly some of the linear algebra work, I think quantum for AI is going to be the last to fall and is very likely 15 to 20 years out.

Yuval: What kind of companies does the Futurum Group work with and what questions are they asking of you?

Bob: We work across the board. The CEO is Daniel Newman, and he started this several years ago and he’s been acquiring a lot of smaller analyst firms. So we work with pretty much all of the large companies that you can imagine. There’s a very large tier of companies below the IBM’s and the Google’s and the Microsoft’s. For example, in data and AI, there’s Databricks, there’s Qlik, there’s quite a few companies like that. HP, of course, is a big company and things like that.

So we do a range of things. We have a very strong media component with them. So this is not the advisory side. This is going to a company and saying, “We can help get your message out much better than you yourself can.” Daniel’s on Bloomberg and CNBC and so forth all the time. Most of us have quite a few followers on social media and things like that, so we can amplify. For the analysis side and the side I’m on, it’s an interesting mix of things. Some people may have seen some of the research notes I put out. All right, well, those are just written. You don’t need a subscription to get to those. You can just go and read them. Those are my thoughts, my views of when something happens, what’s the significance? How does it play into the scheme of things?

But I also work independently with companies who want to have, for example, on the quantum side a view of what’s happening across the industry, and I tend to have opinions about individual companies and how well they’re doing and what they’re doing relative to each other and things like that. So that’s the sort of advice. I will say, if I may, there’s a huge gulf. It’s like if you look at the quantum companies, there are very big companies, IBM, Google, Microsoft, AWS, and so forth. And then there’s like nothing in the middle. Everybody has small companies, right? So startups, maybe a few dozen, maybe a few hundred, that’s it. There’s nothing in the middle. And in the AI world, that middle is very full of what’s happening.

So one of the things I’m very interested in actually doing with the Futurum Group is to help develop additional services targeted more at startups and quantum startups. Obviously, you work for one. I used to work for one. I think they all need better differentiation. A friend of mine said the other day, a lot of their websites look a lot like, one, quantum is good. Quantum will solve a lot of problems. We are the best. Contact us here. So I think as we, now that we’re coming along here, I think it’s time for a freshening up, if you will, for a lot of these quantum startups. They have to do a much better job of differentiating themselves from everybody else. There are ways of doing this. And they have to position themselves in the great bigger picture because they’re not going to become the next Dell or Apple or Oracle, right? Probably the odds are against that. So where will they fit and how? And that’s the sort of advice we can give.

Yuval: Were you surprised by the number of people that want a quantum computer on premises and how much money they’re willing to pay for one?

Bob: A little bit. A little bit. I wrote a piece a couple of months ago called The Case for On-premise Quantum Computers. For a long time, it wasn’t really necessary to have one. You could get one on the cloud. You can go to Amazon, you can go AWS, you can go to Microsoft Azure, you can go to IBM of course for theirs and a few other ones. So if you’re just playing with quantum, you’re experimenting, okay. But if you’re going to use it a fair amount, waiting in a queue for a few hours or for a system that has only a few hours of availability a week is not really the way of doing productive work.

And so for that reason, I think people are investing in these smaller quantum computers, but on-prem if they’re serious about it. Cloud can add up. Cloud sounds great until you get the bill and you realize exactly what they’re charging you for. So I think it’s an interesting balance. Sometimes it’s security that people care about, but they must realize that whether they are purchasing or renting, leasing, whatever on these quantum computers, the quantum computers that will be available a year, two years, three years down the line will be far superior. So they better include some sort of upgrade path into whatever they’re acquiring, right, if they want to put it on-premise.

Yuval: You spoke earlier about partnerships. Some companies believe that in the status of today’s quantum computers, you really need a full-stack optimization. You need to essentially do co-design all the way from the hardware layer to the application. Where do you stand on that as opposed to, we’ll just have a hardware company and maybe a middleware company and maybe an applications company?

Bob: Well I do think we’re very lucky that IBM started Qiskit, Google started Cirq, and there’s Penny Lane, Q# from Microsoft. That is, for a very long time now, and Qiskit started in 2017, we have had a very good stack of software, and the others as well, that have evolved. It started in 2017, but just went to version 1.0 a couple months ago. So I think that has made people’s lives a lot easier. They didn’t have to invent their own software to do it. And here we’re talking about gate and circuit models, analog models are a little different. So that allowed people to focus on their hardware, and they could write backends for something like Qiskit that ran on what they do, Cirq as well. So I think that’s fine, and I think that should continue.

I think we have learned from Linux, even though it’s an operating system, Qiskit’s not an operating system. We have learned the lesson of when software needs to be open, when it needs to be proprietary. And oh, by the way, you have to keep feeding the open source monster. You have to keep giving it more and more code that you originally thought was going to be proprietary. I think a lot of the quantum hardware companies will get more vertically integrated at the level below them, or at least do a lot more to secure their supply chains. For example, in the types of quantum computing systems, so trapped ions, neutral atoms, you need lasers. Go and try to buy a laser. It’s hard. There aren’t that many providers. They’re backed up in terms of orders and things like that. There just aren’t that many providers.

So you asked before, does the government have to step in? Well, what are the essential technologies to make quantum useful and successful? Well, that’s one area, lasers, photonic integrated circuits, and so forth. Without that, you’re going to be dealing with huge optical tables, not exactly the sort of thing you put on a plane or in a data center. So there is a mix of things. I do think there’ll be a little bit of vertical integration. I think there are going to be some tight partnerships of complementary hardware that will eventually lead to mergers of these companies. They’ll try before they buy, so to speak, but then they will kind of come together in a different way. And I hope most of the software is driven through open source because I just think that’s the best way of leveraging what people can do. In particular, they can get a lot of university contribution beyond what they themselves would pay for.

Yuval: As we come close to the end of our conversation today, other than differentiate, what other advice would you give startup CEOs in the quantum space?

Bob: Please watch the hype. We’re all very highly trained to note it very quickly. And if you lose people within the first few seconds by gross hyperbole that you can’t back up by actual experience, people lose interest in what you’re doing. Be honest about your roadmaps. It’s taking longer than we thought. That’s okay. It’s okay to say something will take two years instead of one year. Now I do understand, again, that VCs like a certain amount of time, but there are other ways of getting money and things like this. I said something, I think on LinkedIn earlier in the week, AI is sucking up an awful lot of money out of quantum. So here’s the problem. You have $5 million. Do you want to invest in a quantum hardware company or an AI software company? Well, there are a lot of Silicon Valley types who would tend to go toward the AI company.

Now there can be value, but you have to show that you’re good. You know, as we talked about the roll-ups and the solid partnerships that you are willing to put together the right pieces in a creative way to ultimately have something that’s going to be useful. So look beyond your initial technology and make sure you’re an essential part of something bigger and concrete and stretch your money. That’s always good startup advice, but it’s very true for quantum these days.

Yuval: In last hypothetical, but please, without peaking at what you said in the last two times, she could have dinner with one of the quantum greats, dead or alive, who would that person be?

Bob: Well, yeah, a lot of people say Dirac, but Dirac wasn’t a great conversationalist. Feynman was a great conversationalist. Einstein is kind of the easy choice. I think, you know, it would probably be somebody like Planck. Let’s go way back to 1900, right? The birth of quantum. So when people were first getting that inspiration saying maybe classical physics isn’t quite the way to go. Planck or actually that would be interesting, but Niels Bohr. Niels Bohr was a great man. He was brilliant and he was a wonderful humanitarian. He’s just a great person. So I’ll put him first and then I’ll become more historical with Planck. If I’m lucky, they’ll both be there.

Yuval: Wonderful. Bob, thank you so much for joining me today.

Bob: My pleasure. Thanks for having me back.

To subscribe to the audio podcast, please Spotify here

 

Yuval Boger

Share this article:

Keep track of everything going on in the Quantum Technology Market.

In one place.

Related Articles

Join Our Newsletter