Design News is part of the Informa Markets Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

DesignCon 2019 keynoter Irfan Siddiqi

Everything You Wanted to Know about Quantum Computing

Ahead of his keynote at DesignCon 2019, UC Berkeley physicist Dr. Irfan Siddiqi discusses the realities of quantum computing, where the technology is today, and where it's headed.

Quantum computing is one of those terms that many of us have heard, but not many actually understand. There's been plenty of buzz around quantum technologies recently and investments in the space are at an all-time high—around $177 million, according to analyst estimates. That number is forecast to skyrocket to $15 billion US by 2028.

So why all the excitement? And what is quantum computing exactly? Are we just talking about faster processing and better storage? Dr. Irfan Siddiqi, a professor of physics at the Quantum Nanoscience Laboratory and the Department of Physics at the University of California Berkeley, says that quantum technologies represent a whole new way of thinking about computing.

Dr. Irfan Siddiqi

Ahead of his keynote at DesignCon 2019, Siddiqi spoke with Design News to separate the facts from hype of quantum computing, where the technology is today, and the impact it could have in the future.*

Design News: Let's start at the beginning. What's your definition of quantum computing?

Irfan Siddiqi: For me, any quantum technology, including quantum computing, is something that takes advantage of entanglement. And entanglement is the idea that if you have different pieces of matter and you put them together, they behave as a single unit.

So, for example, each of the bits in a classical computer are independent of each other. If you flip one, it doesn't affect the one next to it. In a quantum computer, all of these bits have correlations with each other, so they're all tied together like one big mass. In fact, the number of states that they can occupy is exponentially larger because of these linkages between neighboring elements.

Quantum computing is the science of manipulating this entangled set of bits for some particular problem of interest in either fundamental science and computation or to do a simulation of the natural world.

DN: So are speed and processing power the only advantages here? Or are we looking at other things as well?

IS: I would say it will actually be a bit deeper than that, although those terms are correct. The power behind quantum is that it changes the types of problems you can solve. So it's not as if it goes a little bit faster than your classical machine. It will take those problems that are impossible on a classical machine and make them possible in real time. And the reason for that is the structure that quantum machines have, where information is stored in such an exponentially larger capacity. It allows you to process things in very different ways. And not only are they faster; They're just really different problems. Problems that are intractable in the classical domain become pliable using quantum devices.

DN: What sorts of problems are those?

IS: The one that is usually given for general computing is factoring a number into its prime factors. This is the hallmark example of quantum computing. Shor's Algorithm is the one that's used to factor a number into its primes. This is the modality of security that is most often used. The ability of classical computers to multiply numbers, but their complete inability to factorize them, is what is the basis of many security algorithms.

But quantum computers can do that very readily. So that's a very good cryptographic, cryptology, communication kind of example.

More on the scientific domain: If we look at even the structure of chemical molecules, even though the theory of quantum mechanics told us about 100 years ago how to solve for their energy structures, we are unable to do this for something beyond an order of about 10 atoms.

So that's kind of shocking—that even with the best computers, you can't solve something more than 10 atoms. There are many, many, many atoms that make up even the simplest of molecules. By using quantum devices, we would be able to, in principle, understand the chemical structure dynamics of things much larger than that number and that could have very profound consequences for artificial solar cells, new types of fertilizers, new types of catalysts—all sorts of things in materials science that, at the moment, are really unreachable to us using classical computations.

DN: So with that, how feasible is it to actually build a quantum computer?

IS: What is involved in building a quantum machine is that we will have to extend the size of our computer bit by bit.

So this sounds a bit strange because we're used to having trillions of transistors in even the simplest of computers. Even your watch probably has many more transistors than that. Nonetheless, one has to sort of go back a bit to the day of vacuum tube computers and even before that, where literally we were putting together computing architectures bit by bit, transistor by transistor. This is the state that we're in with quantum devices.

So it starts off by taking a physical system that exhibits quantum mechanics. At the moment, we have a very preferred physical system for classical computers. This is silicon technology, and we use it to make transistors and so forth. We have not arrived at the one magical technology platform that can be readily used to build a quantum computer. We're still surveying the field and all the sorts of different technologies that could build these devices.

One example would be to use individual atoms or ions, so you literally put a computer together one atom at a time. This would be called an ion-based quantum computer.

We are taking a different approach [at Berkeley]. We use electrical circuits. They are made of metals and each circuit is a resident circuit like an oscillator or a pendulum. Each one of these electrical oscillators is a qubit [quantum bit]. We build these out of superconducting metals, so if we put them to very low temperatures, there's no resistance. It's the resistance that causes the pendulum to slow down. The idea here is that you would build an oscillating circuit that doesn't stop for very many oscillations and allows information to be stored without being lost.

That's one quantum bit built out of electrical circuits using superconducting metals. The goal is to put many of these circuits together and entangle them and build a quantum machine. So the state of the art right now is somewhere between 10 and 100 of these qubits for any modality, whether it's ions or superconductors or what have you.

DN: Is just 10 or 100 qubits enough for any realistic application?

IS: I would say, in the industrial or commercial domain, we have no quantum machine that can outperform a classic one at the moment. This is called quantum supremacy or quantum ascendancy. There's a race to build a machine that can outperform a classical counter.

Here is an example of a multi-qubit chip for computing things, such as quantum simulations of advanced materials, that was created at UC Berkeley. (Image source: Quantum Nanoelectronics Laboratory, UC Berkeley)

DN: Today, we think of processing power in terms of gigahertz and the like. Is there a way to translate that into qubits and draw an equivalency between qubits and, say, transistors today?

IS: There's still a debate in terms of how you benchmark the quality, the processing speed, and power of a quantum device. The reason this becomes a tricky question is that it depends on the algorithm you want to run. Certain architectures, for example, can connect the qubits in a star formation rather than having the nearest neighbors connected. If I arrange qubits into a star, the one in the center can talk to the other ones very easily. If I put them in a ring, I'll have to go around the ring. Or, if I put them in a square block...those all have a different functionality.

Part of this is still evolving. We have to decide what the actual architecture is. Once we agree on the architecture, we can try to benchmark the capability.

However, what one can do is take a certain problem that's difficult on classical computers and see how many qubits you would need to solve that problem. That gives a little bit of a sense of this crossover.

In terms of the chemical spectrum problem, for example: If you take a supercomputer and you try to solve for all of the states in the chemical structure, you'd run out at about 50 or 70. The reason for that is you run out of memory. Because you can see all of these quantum states the more you concatenate this object, it has more and more states. And it grows exponentially, whereas the memory in your computer—even if it's at the petabyte level—just grows linearly with the number of pieces that you have. That crossover is right around 50 or 70 qubits.

What's actually interesting in this area is: Once we've built quantum machines that have, let's say 50, 70, or 100 qubits, they will not be able to be solved or modeled by a classical computer. That is an interesting problem in and of itself. How do you describe a system of so many qubits when you can't do it classically?

DN: In those terms, it sounds like we're talking about a complete paradigm shift. It seems like it is the equivalent of asking how many vacuum tubes a transistor is worth.

IS: Yes, there's no easy way to map this. It's totally shifting the architecture—not only the physical technology, but also the way we store information.

DN: What does that look like then? Does a quantum chip look the same as a chip as we understand it today? Or is it something entirely new?

IS: The qubits that I work on kind of look like a semiconductor chip to the naked eye. It's made on a piece of silicon. It has metal traces. But if you look under a microscope, those metals and wires are arranged in a very different way than a classical transistor. And we have to pay attention to a whole different set of materials problems than silicon folks think about.

DN: What are the materials technologies people are looking at when they research quantum devices?

IS: In my field, we're looking at superconducting metals. The difficulty with this quantum information business is you may ask, 'Why don't we see quantum mechanics every day?' After all, this is the theory that fuels everything, so why isn't my tennis ball, my transistor, etc. quantum? It's all the same stuff, right?

The tricky part is that when quantum systems interact, they entangle. For example, a qubit talks to stuff in its environment. It gets entangled with that. It talks to light that's coming and it gets entangled with that. So this information spreads very rapidly.

It's the idea of decoherence, meaning that quantum information is no longer in your possession. Think about having a little box of information. You have a little wooden box that has a certain amount of information in it. As long as you keep the box closed, then you have it and it's fine. But as soon as you start opening up that box, and different people take out pieces of paper and put in pieces of paper, it's not that the information is destroyed. It's just shared amongst all those different people. The problem is that, until you talk to those people, you don't know what's in that box anymore.

DN: And is that just inherent to the nature of dealing with things at the quantum level?

IS: That's right. Atoms talk with other atoms and things that are around them. So the name of the game is to isolate them in such a way that they don't talk to all these other systems.

At the same time, you—as an experimenter or a user—do want to talk to the computer. So that makes it really hard. It's not as if I can always keep the box closed. Somehow, I have to design a quantum system one step at a time such that I can access it as a user, I can control it, and it doesn't talk to all sorts of things around us.

In the silicon world, silicon is a solid state material. There are ways to think about how to use that same silicon and the same tools that we have in Silicon Valley to produce quantum computers. One approach for doing that is to put in individual implants. So you put phosphorus atoms one at a time into a silicon substrate and use them as your qubits. There are people in the world working on this, but you can imagine the challenge. I would have to control the deposition of a single ion at the nanometer scale one at a time and also bring wires to that ion one at a time. Silicon technology lithography is not there yet. We're not able to control fabrication at the level of one atom at a time. Maybe we can do 10 nanometers or so, but not an atom at a time. So that already makes it difficult.

Microsoft is looking at different types of metallic structures that have some kind of symmetry properties that protect quantum information—for example, if you build a box so that when you open it, it makes another marker someplace else so the information won't go away. Those are called topologically protected qubits, meaning their structure is significant.

Think of a wire that has no twists. Twisting that wire is a way to store a zero and a one. The nice thing about this twist is: No matter how I bend the wire, it still has a twist in it. So this is the idea of using topology—the actual structure—for storing information rather than its state. Microsoft, for example, is working on things of that nature, but they're very far off.

At the moment, we don't have any sort of concrete results in terms of having qubits that function with this type of encoding. Things that are working today are superconducting circuits and trapped ions. With trapped ions, you take a bunch of laser beams, put them together, and that's where an ion can sit. Then you put lots of arrays of these intersecting laser beams and you park ions at each one of these points.

DN: Temperature is also an important factor, right? Do all quantum machines need to be assembled at near absolute zero temperatures?

IS: When your environment is warm, it has vibrations. It has ways for information to leak throughout your system. That's what actually causes decoherence. For superconducting circuits, if you don't go below 1 degree Kelvin—very close to absolute zero—they're not going to actually go into the state where there's no resistance. So that's simply not going to work. Moreover, even if you have semiconductor devices, it's all the other jiggling and vibrations around you that you want to kill at low temperatures. That's why we typically go cryogenic. But it's not a deal breaker at all at the moment.

DN: What about storage? Is there a quantum equivalent to that? Can we imagine a future when people will be trading their mechanical and solid-state hard drives for some sort of quantum drive?

IS: Some of the most challenging problems with this field actually don't have to do with the chip. It's true that we have to make the chip resistant to decoherence and all that, but how do you get all that information onto your chip and off your chip? This is a huge problem.

Let's imagine you're saying that your quantum computer can search a database that has a number of entries equivalent to the number of particles in the universe. How are you going to load all those database entries onto your quantum computer?

There are ways to store quantum information. By the same token, in terms of processing, if you have this entangled structure, you can store a lot more information in a fewer number of particles.

Going forward in quantum technologies, we're going beyond the stage of just thinking about individual qubits or a few qubits in a processor. We need to think about the full structure. What's my memory element? What's my processor element? What's my cache? What's my RAM? All of these ideas are coming forward now.

What's very interesting is that you may use different materials systems for different functions. Some systems that interact very strongly with their environment can be used for the processor. So superconducting qubits may be good for that. Atoms may be great for memories. You shove them away or right into some system or arrangement and they sit there for a long period of time. There are very sort of cutting-edge ideas around how to actually go from individual qubits to an actual architecture.

DN: Part of your work at UC Berkeley has been establishing an Advanced Quantum Testbed (AQT) at the Berkeley National Laboratory. Can you talk a bit more about this and what it means for the field as a whole?

IS: Progress in this field depends on advancements in many different places. For a long time, most of the investments and advancements in this field were coming from academia. So you had professors thinking about what a quantum device is and demonstrating quantum devices.

Once we get to larger numbers of qubits, the question becomes: What's the most efficient way to actually build a machine? One way is to say, 'Well, I'm going to open up companies, and those companies will go ahead and try to build these devices.' This is great. The challenge of it is that we haven't agreed on the fundamentals yet. We don't know what the right material is to do this. We don't know the right architecture. We don't even know the right algorithm to run on these things.

Dr. Siddiqi (center) works with other members at the Advanced Quantum Testbed at Berkeley Lab. (Image source: Peter DaSilva/Berkeley Lab)

The AQT is one example of what a quantum machine could look like. And it's not obvious that that will actually scale to the ultimate device. Let's imagine a company goes ahead and builds ten and a hundred qubits. That's great. I think the great value is that we're learning how hard the problem is. When you're learning about the things you need to solve in these early days—and what that immediately motivates—it's very efficient to have an academic-like entity that can work with much larger numbers of qubits in a very risky environment.

The AQT is the first of these sort of national lab ventures to build a quantum computer of a certain scale that's open. It's public. People can come use it and partner with it. We learn a lot about quantum science this way.

I think it's beneficial to industry as well. It's not as if we're doing something that's competitive with them. This is all pre-competitive research. In fact, we are trying to answer those hard questions, which will guide companies in the future in actually building mature quantum technologies.

DN: We've talked a lot about the hardware end of quantum computing. But what about the software side? Are developers going to need to learn entirely new programming languages to interface with quantum machines?

IS: There are a lot of people looking at the entire software stack at the moment. Another advantage of this type of testbed is that you can look at that programming stack from the machine language level—all the way down to where you are controlling the individual operations, all the way up to the compiler level or even beyond that, which is the user interface—and say, 'Here is a program I want to run.' But how you interface all of those pieces together in some efficient way is quite a challenge. Many people are working on different pieces of this puzzle—the software stack and firmware stack. What we can do with this testbed is integrate all of these pieces together and learn what the issues are.

DN: So is it feasible to think that someday, a programmer with a background in say, Python, will find that their knowledge applies in quantum computing?

IS: Actually, we use Python quite a lot! Remember, at the end of the day, there's a quantum processor that's there. But at some point, you have to interface with the classical world. That classical world is still Python and C and so forth.

In our current implementations, the core is this chip that has a certain number of 10 or 100 qubits and pulses that you send to them. How to coordinate all those pulses and the readout out of those pulses is all coordinated by machines that are running Verilog VHDL Python at that level.

Hardware control is still classical. But then you go into different flavors of compilers and things and think about things like, 'How do I lay up my gates?' That's different than your classical gate set because the logic is different. So those compilers and optimizers are necessarily different because of trying to tie together a different gate set. Then you go into optimizers and, finally, something that translates your high level circuit. At the moment, I have If, Then, Else, Boolean logic, and other things that I write for my program...I have to write that in quantum's language now.

You need an interface to put that into the system and then finally translate that into your native gate set to your hardware and execute it. We do have examples of all of these different pieces, so Microsoft has some languages, Google has languages—the folks trying to write converters that go from a circuit that a physicist or an engineer can write on the board to something that would turn into a gate set.

But again, not having all the hardware and software co-located in some open fashion means that people are developing entities without actually having a means to test them. And I think that's where the testbed really comes in. If you develop a better programming language, a better compiler, then we have a technology cycle every year. We're going to look at different proposals to improve the technology on the testbed to allow people that actually don't have this heavy duty hardware in their backyard to go ahead and implement.

DN: How do you do code like If/Then when you're talking about superposition and it's both at the same time?

IS: The idea is: In a quantum system, all of your operations have to deal with the system as a whole. So the first thing I do is generate entanglement. Then, I will have a series of operations that take this entanglement into account. It's fine for them to be entangled and superimposed and all that. These very special sets of rotations are all going to move together. When I flip one, it's going to flip another, and so on. They're going to move together. Then, at the end of the day, I'm going to measure one.

So basically, it's only at the very end that you actually get a definite answer. Everything in between has to take into account this sort of measurement—this sort of entanglement that's there.

Let's imagine that you are running a race with three people and you're all holding each other's hands. Say the problem is that I have to weigh one of you. I can't weigh you when all three of you are together. So at some point, I'm going to peel off one of you and weigh you. Basically, you are going to keep that entanglement and correlational alive until you know you're done. Then, you do the final read.

DN: Obviously, we're a long way from seeing quantum computers in retail stores. But what's really the very logical practical next step right now?

IS: I think the most important thing to do is to figure out what these machines may be good for. For example, if you are in the vacuum tube era and you said, 'I want to make a zillion of these vacuum tubes and have them live for weeks or years,' you would say this is an impossible problem.

I think in this era, where we're still playing around with a vacuum tube equivalence, it's important to figure out: What are the technologies that are robust and allow us to process quantum information? What are the algorithms that can solve problems of greatest impact? We have a few.

One has to figure out what those problems are and then figure out what the classical technologies are that we also need to develop to interface with the quantum core. How do we actually get information on and off these things?

The real value in this first five or 10 year period is to figure out what quantum technologies are good for and how you get there—rather than saying, 'By next year, I want to have the machine that's going to be my supercomputer.' That's unlikely unless we have a breakthrough.

In the path of actually moving forward toward an engineering route to quantum technologies, we are at a critical phase where we're able to integrate the little pieces together into a full machine. Before, we had little pieces. Somebody made a qubit, somebody made a memory, somebody made a gate, somebody wrote an algorithm...And it wasn't obvious what this does. So we're at the point where we can say, 'Let's kluge all of this stuff together and build the machine.'

But to say that the machine is the final architecture is not really what we're talking about. The most useful thing about it is to figure out how to actually turn this prototype into a product. And I would bet money, or at least I have great hope, that the real applications where this is going to take off are still to come. It was the same with the classical computers. We didn't have Facebook in the era of ENIAC. Once you actually build the hardware and it starts to do things in an efficient manner, that naturally motivates the application, which will bring this thing into more general use.

Now, mind you, sometimes one asks, 'What about the cryogenic aspect of it?' But maybe not everyone needs to have the physical computer on their desktop. There was a long era when we had computers at terminals, so that you can access such computing resources through the cloud. When Google is searching something or Facebook is doing something, I only have the app. I only have that sort of interface on my local machine. I don't actually have the hardware server right in my backyard. It's no different with quantum. You can imagine a certain number of quantum computers existing in some exotic environments that are remotely accessible.

*This interview has been edited for clarity.

DesignCon 2019 engineering education By Engineers, for Engineers
 Join our in-depth conference program with over 100 technical paper sessions, panels, and tutorials spanning 15 tracks. Learn more: DesignCon. Jan. 29-31, 2019, in Santa Clara, CA. Register to attend, hosted by Design News’ parent company UBM.

Chris Wiltz is a Senior Editor at  Design News covering emerging technologies including AI, VR/AR, and robotics.

Hide comments
account-default-image

Comments

  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.
Publish