The computing world was abuzz last week after Google scientists announced they’d passed a key threshold in which an experimental quantum computer solved a problem in just minutes that a classical computer would take years —10,000 by Google’s count — to solve. The pronouncement, later disputed by scientists at rival IBM, was widely hailed as proof that quantum computers, which use the mysterious properties of matter at extremely small scales to greatly advance processing power, can — as theorized — exhibit “quantum supremacy” and vastly outperform even the world’s most powerful classical computers.
At Harvard, George Vasmer Leverett Professor of Physics Mikhail Lukin watched the announcement with interest, in part because he — together with collaborators from Harvard and MIT — already built a quantum machine of similar power to Google’s and used it to solve scientific problems. Lukin, who co-directs the Harvard Quantum Initiative, spoke with the Gazette about the week’s quantum computing news.
GAZETTE: What is “quantum supremacy” and why is it important in talking about quantum computing?
LUKIN: Let me first describe quantum computers. They constitute a new approach to processing information, one that makes use of the laws of quantum mechanics, the discipline of physics that describes the behavior of particles at a microscopic level, of atoms, of nuclei.
GAZETTE: Does it take advantage of differences in behavior at those very tiny scales from what we might expect in the macro world?
LUKIN: It takes advantage of differences in behavior and in particular one very strange feature of the quantum world. That is that objects can be in several different states — in several different places — at once. That sounds very bizarre, but in the quantum world, an object can be in your office and in my office at the same time. Even though it sounds strange, this idea of quantum “superposition” has been confirmed and it’s been routinely studied in experiments involving microscopic objects like single atoms, for example, over the past century.
GAZETTE: Now, this is not a new idea, or even — at this point — a revolutionary idea because it’s actually used in certain applications that people see every day, right?
LUKIN: Yes. Even certain technologies such as magnetic resonance imaging (MRI) are based on this idea of superposition. So when you get an MRI in the hospital, superposition is being used. Superposition seems bizarre because these effects do not occur in the macroscopic world. Quantum superpositions of microscopic objects are extremely fragile and susceptible to any kind of environmental perturbation. A single photon hitting quantum superposition can cause it to collapse. That means when you look at it, you will always find an object in one place or another. The idea of quantum computing is to make use of these superpositions for massively parallel processing of information. If you were to use a classical computer, you’d code your information in a string of zeroes and ones. In a quantum computer, you can prepare a state that has all sorts of combinations of zeros and ones, it can be in one state, and in another state, and in yet another state all at once. Then, as long as it’s quantum mechanical, as long as it can preserve this superposition, it can process all of these inputs simultaneously. And that massive parallelism gives rise to a very powerful computer.
GAZETTE: What is the threshold of “quantum supremacy” that Google said it passed last week?
LUKIN: This massive parallelism enables one to exponentially speed up quantum computation over classical computers. Quantum bits are the analog of bits in a classical computer, and are used to store a superposition state. So, even if a quantum computer has just 50 quantum bits, which seems very small, it is very challenging for a classical machine to simulate it. And the reason is that even 50 qubits in a superposition state can store and process exponentially many more combinations at once. If you had a system of 300 qubits, you could store and process more bits of information than the number of particles in the universe. So this idea of supremacy is that you build a system large enough, quantum enough, and programmable enough that you can execute operations that the best possible classical computer just cannot simulate in any kind of reasonable time.
GAZETTE: In this specific case we’re talking about a quantum computer with 53 qubits. How big is it physically? How would it compare to, say, a supercomputer?
LUKIN: In terms of physical size? It is a room full of equipment. Something like the Summit supercomputer in Oak Ridge National Lab. It’s probably comparable.
GAZETTE: But it has exponentially more computing power?
LUKIN: That’s the hope. If you have a 50-qubit system and run it long enough to execute a general enough algorithm, it will be very, very challenging for the best classical computers to catch up. And if they can catch up, you could just add a few more qubits — above 60, 70, or 100 — and it’s very clear that it will be completely impossible for a classical computer to catch up.
GAZETTE: IBM, which is a competitor, cast doubt on Google’s achievement. Do you have a sense as to who’s right or who’s wrong here?
LUKIN: Google’s achievement is quite impressive. But it points to one specific calculation and says, “We crossed the threshold.” I think, in practice, it’s not quite like that. There is no doubt that once quantum computers become large enough, classical computers cannot simulate them. That’s very clear. It is also clear that 50 qubits is a sort of threshold and the system that they built is very competitive with the best systems that exist around in the world, including one here at Harvard. What they’ve done is a great example of how to test for this so-called supremacy idea. However, 50 or so qubit systems already have been used in other labs, including ours, for the past two or so years. There have been a number of experiments done with systems of that scale which classical computers have a hard time catching up to. In this sense, I would describe what’s going on now as not a singular event but more like a wave that is coming. It might be that IBM folks found an algorithm to efficiently simulate on a classical computer what the Google quantum computer did. I’m not surprised by that, but at the same time, it’s very, very clear that we’re entering, as a community, a place where no one has ever been before, meaning we can do things much faster than the classical computers. It is actually happening.
GAZETTE: What was the problem that Google was trying to solve?
LUKIN: The specific problem that Google is trying to solve is akin to generating random numbers in a quantum way. The algorithm they’re using is not designed to be practically useful. So in this practical sense, quantum supremacy by itself, to me, does not mean very much. But what will be really exciting — and this is a key goal in the field — is achieving that quantum advantage for things that are useful. You execute the algorithm and actually learn something. There are two types of useful quantum advantage, one for scientific applications and the other for general purpose applications. Google’s paper is something in between.