- R&D 100 Awards
Even for advanced programmers, quantum computing—which is based on the principles of quantum mechanics—is not an easy concept to wrap your head around.
So much in fact, that even Albert Einstein was skeptical of the theory of quantum mechanics for years, said Robert Sutor, PhD, vice president of IBM Q Strategy and Ecosystem, IBM Research.
“One of the two properties that is often spoke about with quantum mechanics, and therefore quantum computing, is called entanglement,” explained Sutor. “It is two particles that are somehow so linked that if you know something about one, you know everything about the other. The theory says that even if those two things were hundreds of light years away that would still be true.”
“Einstein did not like this idea. He called it, ‘spooky action at a distance.’ He thought it was just ridiculous,” said Sutor. “He eventually was convinced otherwise. But, if Einstein took a few years to accept something, maybe some other people might too.”
That is why IBM has released the IBM Quantum (IBM Q) Experience, a one-of-a-kind online platform that enables scientists and science enthusiasts all over the world to explore IBM’s emerging quantum computing hardware from home. Already more than 83,000 people have run over four million executions using IBM’s five and 16 qubit quantum computers.
The IBM Q was also a 2017 R&D 100 Award winner, receiving the award Nov. 17, 2017 at the black-tie R&D 100 Awards gala in Orlando, FL.
The idea behind the early release of the IBM Q was to give today’s researchers, programmers, students and scientists a head-start on learning the complexities of quantum computing before the technology advances further, said Sutor.
“With the Q experience we can demonstrate quantum entanglement using very, very simple circuits with two or three qubits,” said Sutor. “You can sit at home and you can show actual quantum containment. Einstein didn’t have that advantage.”
But as of now, IBM Q demonstrates only a tiny fraction of what could be possible with quantum computing, said Sutor. In an interviewer with R&D Magazine, Sutor explains exactly what quantum computing is, its future potential, and why IBM decided to release it to the public now.
R&D Magazine: What is quantum computing?
Sutor: The computing that people are used to right now, in their laptop, their cell phone anything like that, is something we call classical computing. That technology goes way back to the 1940s—with the very earliest types of computers—and it has served us very well. However, over the years, the question has become—can classical computers keep up with what seems to be our insatiable desire and need for additional computer power? Evidence of the idea that you can’t just have one chip that keeps getting faster and faster, was seen when, a few years ago, computers and then game consoles and even your smartphone, started to contain multiple cores. Gaming consoles now have eight. If you can’t make a single processor faster, you have to keep adding more processors and then link them together, but there is a limit to how much you can do. With this in mind, people started asking, is this the only way of doing computing? This type of classical computing really isn't very good at modeling.
When you are using classical computers, even high performance computers, to try and compute within chemistry—the interactions between atoms, the electrons, the energy and all of that— it is very hard to do, and you really can’t do it terribly accurately with a classical computer. Once these molecules get beyond the very small, you just run out of capacity and you have to start approximating everywhere. Of course, when you do approximations it is not exact.
The question was posed by a professor named Richard Feynman in 1981 saying, 'when we think of nature and we think of how nature works, to the best of our knowledge, the thing that seems to describe it is quantum mechanics.' Quantum mechanics has a long history and there were a lot of people that didn’t believe at first, but through the years, as it was developed, it was pretty much decided that it was the best thing that we know about that describes the theories of the very small.
What Feynman said was, instead of basing things on our classical computing, which we used to call digital computer technology, if we had computers that used the same principals of quantum mechanics, we could therefore map these more nature processes, like chemistry, over to such computers, and then work with them much more easily and far more accurately and exactly. Hence there was this notion of going from quantum mechanics to quantum computing.
R&D Magazine: Why do we need quantum computing? How can it be applied?
Sutor: The example I like to give to show why you might need this type of thing is the caffeine molecule. Caffeine is not a very large molecule. It has 95 electrons, and lots of carbons and hydrogens, and so forth. If you were to take a single molecule of caffeine and freeze it at one instant in time, the amount of information you would need to describe its energy configuration exactly is 10^48 bits. That’s a lot. For comparison the number of atoms in the Earth is estimated to be between 10^49 and 10^50 bits. So you would need a number of bits to describe one single molecule at one single instant comparable to 1 to 10 percent of all the atoms on the planet. That is not going to happen.
Even this very small, very well-known model, you are never going to be able to represent this exactly using classical computing technology. Whereas on quantum computers, we don’t use bits as the basic units of information, we use something called qubits, which have these strange quantum mechanical properties. When you compute with them, they give you access to a tremendous amount of working memory. So much so, that the caffeine molecule, which was impossible to represent classically, you could represent exactly using 160 qubits.
Today the largest quantum computer we have discussed is 50 qubits—we have a working prototype for that. So, hoping it all goes well over the next few years, we can say there is a reasonable possibility that we will get to 160 qubits, and of course we’d love to get far beyond that.
So we go from this situation that is completely impossible on the classical side, but using a very different information model and way of computing with qubits and quantum computing, we can see, in what we hope is the not too distant future, a way of representing such molecules. Once we get additional power—more qubit that behave well, that don’t have too many errors in them—then we will be able to manipulate such molecules and much larger molecules as well.
From there we branch off. Quantum computing seems to be able to represent a lot of information, but the fundamental principles are different. So could we apply this to other areas? People are looking at some types of optimization problems. Financial services problems, for example, where you are trying to minimize risk in a portfolio and things are constantly changing and there are lot of variables and a lot of combinations. Might it be possible at some point to use quantum computing to significantly speed up and make more accurate those types of computations?
There is also some potential in artificial intelligence. Quantum computing is not a big data technology; you do not push a lot of data through a quantum computer. But down at the heart of a lot of AI solutions, you do math, you run algorithms. So it is reasonable to say that there could be some areas of quantum computing that could significantly increase the speed or the power or even make things possible that would otherwise not be possible.
R&D Magazine: What applications could quantum computing have in science?
Sutor: Quantum computing could potentially be used for the discovery of new materials—people are always trying to discover new alloys or new physical materials that have particular physical processes. It’s a fun thought to think, could quantum computing be used to develop new materials to build better quantum computers? We don’t know.
Another area of potential is drug discovery. Now I do really want to caution that this is a long way off. As I mentioned, we aren’t really there with caffeine, and the molecules required in pharmaceuticals are much larger and much more complicated. We are going to need tens of thousands more high quality qubits to really do interesting things there. But when we get to that point, we may even change the phrase. We may go from, drug discovery, to drug computation. Which means, don’t find something that you think will work, actually compute what it is you need. You could say, 'I need a molecule, a drug, which has certain properties,' and then compute what that drug would be. That is a ways off, more than ten years. But it opens up interesting areas. But I do want to stress that these are all areas people are exploring but none are guaranteed.
R&D Magazine: What do you see as the role of the IBM Q Experience in furthering the field of quantum computing?
Sutor: In 2016, we at IBM research had gotten to the point where we could build, what we consider, reliable five qubit computers. Five is very small number of qubits and we had to think if we should just keep this to ourselves and keep working. But we made this decision that we were going to make this machine available on the web. What that means is that for the first time, the general public, could go to a website—it’s called the IBM Q experience—do some programing and do real, actual quantum computing. It wasn’t just a simulation, it allowed people to use quantum computers for the very first time outside of a research laboratory.
People ask the question, 'can I do anything fascinating with 5 qubits?' The answer is yes, you can learn about quantum computing. It is so radically different. The best software engineer in the world who understands classical computers, he or she is probably terrible at quantum computing. We felt that this was a very early point and we could put this out there and people could think about the model, how programs are built, a quantum algorithm, how you would use quantum algorithm, and possible application areas.
We moved from that size in 2017 when we made a 16 qubit machine available and at this point we have two five qubit machines, one 16 qubit machine and software simulators available as part of the IBQ Experience. To date, we have had more than 83,000 people that have used this, and they have run over four million executions. We have gone from really, two years ago, where the number of people in the world who had maybe done some quantum computing was in the dozens, to this point less than two years later over 83,000 people have done it.
We also have a premium program called the IBM Q network. Those are the people that who will get access to the latest and greatest technology. In that program we have 20 qubit machines and we will be putting online what is now a prototype 50 qubit machine. With that network we have individual partners, like JPMorgan Chase. We also have hubs—including Oak Ridge National Laboratory, University of Oxford, and University of Melbourne in Australia—that are centers of excellence around quantum computing. They will receive their quantum computing capacity from us and they will then work with others as well.
Three weeks ago we announced a new startup program. We went out to Silicon Valley and announced that eight of what we think are the very best startups in this area, have joined in with us to really look at advancing the technology from the hardware to the software to the applications.
We also consider education partnerships as part of the Q network. We’ve been working with MIT on a series of classes about quantum computing. There are other universities that we will be working with as well. One target for me when I think about educating people is that I want to get students as reasonably young as possible. For sure in college, but maybe even those high school juniors and seniors. I’ve been coding for 45 years. I have a lot of things burned into my brain as to how one does programing. But if we can start this new generation of quantum-native, born-on-quantum people, that really have some of their first programing experiences on quantum computers—and then also understand how classical computers work with them—these are really going to be the powerful and useful coders in the next few years.
R&D Magazine: What was the motivation for releasing this technology to the public this early?
Sutor: There are no commercial applications out there for quantum computers today. We hope there will be soon. We will need more qubit, we need more high quality qubits. We are hoping that all this early work now with those 83,000 plus users will lead to breakthroughs that will give us what we call quantum advantage. That is a phrase we use that says, we can look in any particular area and say what we just did there with quantum computing is a significant improvement over anything we know how to do classically. That is what we are looking for. What people are doing in the IBM Q Experience is laying the groundwork for that. That is what we want. We didn’t want to keep all these machines internal to ourselves and then at some point a few years from now spring it on people, and say 'congratulations, go figure it out.' With something that is as ‘weird’ as this, there is a lot of education that is needed.
R&D Magazine: The IBM Q is still in the early stages of development, what are the immediate goals for its future?
Sutor: Scientifically our goal is improving what we call the quantum volume. The power of a quantum computer is not simply how many qubits you have—we cannot just keep adding more and more qubits to get more power, that’s now how things work. The reason is this. In classical computers if you set something to zero, it stays zero. If you set something to one, it stays one. This is what is called fault tolerance. What we ultimately need to get to is fault tolerance for quantum computers. The first step to do that is to make the individual qubits be available for use for a longer period of time, that is called coherence time, and also to have them operate with very few errors.
There is a certain amount of time you have to work with qubits and then they basically start wandering off. Clearly, an important goal is to make that coherence time as long as possible. Right now, we are looking at values around 100 microseconds, which is 100 million of second. The longer you have that, the more things you can do with those qubits— that is you can compute with them, you can have longer algorithms and things like that.
The other thing is just general errors. Qubits are in general, very, very sensitive. If anything goes wrong it could throw off the expected value. This has to be considered across a lot of different dimensions that occur in quantum computers. That is, a combination of the materials, the physical layout of the devices, and the software as well.
Basically, you wrap all of these things together, you wrap the number of qubits together, you wrap in the error rate, you wrap in the coherence time, and through the use of a formula you get quantum volume—a number. It is through that number that you can compare the error of one quantum computer to another. Improving that quantum volume is the thing that we are doing.
This interview has been edited for length and clarity