By: Jon Farrow
29 Oct, 2019
Quantum supremacy: Giant leap, or one small step?
October 29, 2019
An array of Sycamore chips being prepared for preliminary electrical testing. (Google)
With much hype surrounding Google’s recent announcement of “quantum supremacy,” CIFAR fellows weigh in on this milestone in computing.
A team from Google, including CIFAR associate fellow David Bacon, published a paper in Nature claiming they had built a superconducting quantum computer able to perform calculations in minutes that would take a normal computer, also known as a classical computer, thousands of years. This type of quantum computer was first proposed by Alexandre Blais, also a CIFAR fellow.
“Quantum Supremacy” is a term coined by CIFAR advisor John Preskill in 2012. He defines it as “the point where quantum computers can do things that classical computers can’t, regardless of whether those tasks are useful.” This has since become a major goal of quantum computing projects at Google, IBM, Microsoft, Intel, Alibaba, and many other companies, startups, and research labs.
Classical computers use classical bits, which can be either 0s or 1s, in order to perform calculations. Quantum computers, on the other hand, use quantum bits (qubits) which can exist in states between 0 and 1. Many believe that problems like cracking certain types of cryptography would be much faster with a quantum computer than with a classical computer. However, the difficult engineering challenges associated with building a real quantum computer mean that the current goal is a proof of concept: to solve any problem faster than a classical computer could.
The Google team’s accomplishment is being hailed by some as a landmark achievement in the history of computing. Others, including Jay Gambetta, a CIFAR associate fellow and vice president, Quantum Computing at IBM, have argued that the milestone of supremacy hasn’t been reached yet. He and his team at IBM claim that a classical supercomputer IBM developed, Summit, could complete the task accomplished by Google in a matter of days.
Scott Aaronson, a former associate fellow in the CIFAR Quantum Information Science program, came up with the mathematical techniques that allow the demonstration of superiority. He described the difficulty of identifying a moment of quantum supremacy in his blog:
“From the beginning, it was clear that quantum supremacy would not be a milestone like the moon landing—something that’s achieved in a moment, and is then clear to everyone for all time. It would be more like eradicating measles: it could be achieved, then temporarily unachieved, then re-achieved. For by definition, quantum supremacy all about beating something—namely, classical computation—and the latter can, at least for a while, fight back.”
We spoke with David Poulin (University of Sherbrooke) and Aephraim Steinberg (University of Toronto), the co-directors of the CIFAR Quantum Information Science program, as well as David Gosset, a fellow in the program and professor at the Institute for Quantum Computing at the University of Waterloo, to learn more about this discovery and its implications for the field of quantum computing. The interviews have been edited for length and clarity.
“In this system, the qubits look like an ordinary electric circuit board, but are engineered with superconducting materials so that quantum mechanical effects are taking place.”
What did the Google team do?
David Gosset: They built a processor with 53 functioning qubits, where they could apply gates or operations between nearest neighbours on a grid in two dimensions. They chose a random computation with this number of qubits and a reasonable depth, then they measured every qubit at the output.
What they get is a bit string, a sequence of zeros and ones. But there are some bit strings that are more likely than others. When they repeatedly ran this random quantum circuit and measured it, they got more of the bit strings that appear with higher probability than the ones that appear with lower probability. And that’s the thing that’s supposed to be hard classically – getting the output bit strings to appear in a way that’s correlated with the specific random choice of circuit.
David Poulin: Google used a superconducting quantum computer (the architecture of which, by the way, was proposed by Alexandre Blais, a member of the CIFAR program), which is one of the architectures pursued by industrial groups. In this system, the qubits look like an ordinary electric circuit board, but are engineered with superconducting materials so that quantum mechanical effects are taking place.
There are other qubits you could use though, like ion traps, photons, Majorana wires, quantum dots, or spin qubits. In the CIFAR Quantum Information Science program, we believe that it’s too early from a scientific point of view to pick one of these technologies, and that each of them has something to learn from the other. So we have members investigating all of these.
“The drive to supremacy is a little bit of a PR exploit, and that’s not to say there isn’t solid research here.”
Aephraim Steinberg: There are critics out there who understand perfectly well the theory of quantum computing and simply said that for physical reasons, once we tried to build large enough systems for them to be competitive, physics would not permit them to continue with this exponential advantage. Now even those people are saying, “this is the scale where if it still works, I will admit I’m wrong.”
The drive to supremacy is a little bit of a PR exploit, and that’s not to say there isn’t solid research here. People identified one milestone, and then they made engineering choices that were not driven by building the best computer in the long run, but driven by proving this particular, narrow kind of superiority. And that’s fine. But on its own, not only does this solve an artificial problem we don’t particularly care about the answer to, but it also solves this problem in the presence of so much noise that even if you wanted the answer, you’d have trouble conceiving of this being a useful way of approaching it.
If we really want to start doing useful computations, we have to learn how to make fault-tolerant systems where you’d be able to keep making the system larger and have it give you an accurate result.
What does this mean for the field of quantum computing?
David Poulin: This is definitely an important step. From a fundamental scientific point of view, it is pretty exciting. Quantum mechanics is being used in a highly controlled way to achieve some computational task. Even though the task does not have commercial use, they have achieved something that would have required extraordinarily more classical resources.
“From a fundamental scientific point of view, it is pretty exciting. Quantum mechanics is being used in a highly controlled way to achieve some computational task.”
We’re used to thinking that quantum mechanics is confined to the atomic world: it is the theory that describes molecules, atoms, and subatomic particles. But here we have an engineered, macroscopic system behaving quantum mechanically. This is pushing quantum mechanics in a regime of highly complex systems that has never been observed before. This also demonstrates the steady and spectacular improvement of the hardware.
Aephraim Steinberg: I think it is a huge milestone. We are now living at this time where quantum computers are finally starting to become more powerful than classical computers.
Asking exactly what instant that was, and whether you should say it only counts if it’s faster than the biggest supercomputer we have, running for more than a year, or whether running for two and a half days is already hard enough, that’s all secondary. The big deal is how this will scale as problems get bigger and more difficult. This is one of the first convincing demonstrations of a potentially scalable architecture. Really tackling a problem large enough that it is nearly intractable on classical devices.
“Asking exactly what instant that was, and whether you should say it only counts if it’s faster than the biggest supercomputer we have, running for more than a year, or whether running for two and a half days is already hard enough, that’s all secondary.”
David Gosset: Up until very recently, there has been a focus in the theory of quantum computing on devising algorithms for a future fault-tolerant quantum computer. We’ve come to a point where now it seems possible to build noisy, small quantum computers. But that is not something where we have a large set of algorithms that are just ready to go. The quantum supremacy milestone really should be thought of as sitting in this noisy intermediate scale era.
These computers, like the Google one, don’t have error correction or fault-tolerance. In the long run, error correction must happen. There doesn’t appear to be a way in which quantum computers will scale without that.
What is CIFAR’s role in quantum computing?
David Gosset: The CIFAR program in Quantum Information Science has a long history, and its membership has included many of the pioneers of quantum information science. I would say that a lot of the way that the field has been formed comes through CIFAR one way or another.
For example, Jay Gambetta, an associate fellow of the current program—my former manager at IBM—is a driving force behind IBM’s quantum efforts. Scott Aaronson wrote what sounds like an obscure paper about the power of post-selection in quantum computing and complexity theory, but the line of argument used in that paper later formed a major component of these quantum supremacy proposals.
“…a lot of the way that the field has been formed comes through CIFAR one way or another.”
If you look at the history, what has often happened is that there is a brilliant idea and lines of work surrounding it that eventually feed into big discoveries. This idea, that sampling the output of a quantum computer is hard to classically simulate, had a seed a long time ago, and then people pursued it for a long time before it turned into this. We need more that: more of these sorts of seminal ideas that take time to incubate.
And most industry is not set up for incubating ideas for 10 or 15 years, or to set the wheels rolling for something that happens down the road.
Aephraim Steinberg: It is hugely important to have organizations like CIFAR that are keeping their focus on the long term scientific goals, that may not make the same press. When someone says, “we’ve developed one logical qubit in a fault-tolerant way,” that’s not going to have the same impact for the public that the present announcement does, but in the long term that could be the more important development.
“It is hugely important to have organizations like CIFAR that are keeping their focus on the long term scientific goals, that may not make the same press.”
We must keep an independent, long term, and fundamental vision, no matter what else is going on in the short term. People have expressed concern that when billions of dollars are being invested in a problem the role for a small group of people is diminished. Actually, that’s exactly why you need a group like this, because there are large sums of money, and the industrial mindset can easily lead you into a kind of tunnel vision. It is crucial to have, in parallel, a group that is insisting on being skeptical and looking at the broad questions. All of the industrial groups that agreed to support or partner with the CIFAR Quantum Information Science program specifically said they knew they needed this. They need this community to engage with. I think that’s a fantastic role for CIFAR to be in, especially during this new growth period.
David Poulin: For the past two renewals of the program, five years and 10 years ago, we said that the main goal was to understand what can be done with first generation quantum computers. This has now been achieved. Now, we want to set the targets for the next 10 or 15 years.
The landscape has changed tremendously and in particular the growing industrial input has had an important impact on the field. What you lose in these growth periods as an individual is the ability to maintain a global vision of the field. And a global vision is very important if you want to make scientific leaps and help shape the future of the field.
With that in mind, we designed a CIFAR program that would be the venue for the leaders of the field to regularly meet and share their vision of the field, identifying key problems and research areas. In particular, the program brings together researchers utilizing a wide range of qubit architectures together with computer scientists and theoretical physicists. Only with such a broad expertise can we stay connected to all major aspects of quantum information science research, and continue to shape the future of the field.
Related Articles
-
CIFAR Roundtable on Global Quantum Strategies
December 13, 2021
-
A Quantum Revolution: Report on Global Policies for Quantum Technology
April 07, 2021
-
CIFAR researchers measure the duration of quantum tunnelling
July 22, 2020
-
CIFAR fellows awarded 2019 Micius Quantum Prize
May 08, 2019