Researchers and tech companies are in a global race to develop quantum computers that can solve hard scientific problems that conventional computers can’t, and that they hope can eventually support advances in areas like drug discovery, AI, and cryptography. To learn more about quantum computing and progress in the field, writer Sara Frueh chatted with Mark Horowitz, professor of electrical engineering and computer science and Yahoo! Founders Professor in the School of Engineering at Stanford University. Horowitz chaired a 2019 National Academies report on quantum computing.

For people who are new to this topic, what is quantum computing? What makes a quantum computer different from the conventional kind?

Horowitz: To answer that, you first need to ask, what is a computer? A computer is a device that stores information as a set of ones and zeros. It’s called a binary bit, and those binary ones and zeros can be put together to represent numbers, colors on the screen, words. So, we represent everything in terms of these bits.   The computer can do only a few operations on these bits — it can add them, and do logical operations between the bits — but we figured out how to convert those kinds of operations, through the use of programs, to do amazing things. It’s the reason your cell phone works. It’s the reason Zoom works and you can see people remotely. We get really nice graphics and games. So, all of that is done through conventional computers that use binary representations of ones and zeros.   A quantum computer is also a kind of computing device, but instead of using a bit — a one or zero — it uses this thing called a qubit, or quantum bit. And the quantum bit has some additional properties that conventional computing bits don’t have.   I think an analogy might be useful here. A bit is like a coin sitting on a table. Let’s say heads represents a one, and tails represents a zero. When you look at a bit, it will be either a one or a zero. A qubit is more like a pool ball with a number on it. When the number is facing up, the qubit is a one, and when the number is facing the table, it is a zero. But since it is a ball, the number could point to the side. In this position, the qubit will be in a “superposition” of states — it can be both one and zero simultaneously.

The fact that qubits can be both zero and one at the same time — and another quantum property called entanglement — allows qubits to do some operations that are hard for conventional computers to do. And there are certain tasks that we know quantum computers can do much more quickly than regular computers. That is the good news. The bad news is that building a machine that can control ball-like qubits is much harder than building a machine that needs to store coin-like bits.   For example, while conventional computers can have millions or billions of bits stored, quantum computers can only have a small number — around 100 — right now. People are trying to ramp that up. And normal computers are very reliable: When you do an operation, it’s really, really improbable that you will make an error. Storing balls and rotating them precisely is much harder, so right now, in quantum computing, errors are not that infrequent.

Why are people so excited about the prospect of quantum computers? What kinds of problems will they be able to solve?

Horowitz: What’s exciting is that there are a number of scientific problems that quantum computers could be more apt to solve, and they potentially could solve some hard problems more quickly. For example, there are questions that people are interested in related to chemistry that depend on the interactions between atoms. These quantum interactions could be solved directly on a quantum computer, which would be awesome. We wouldn’t have to settle for having some approximate solutions the way we do now when we use conventional computers.   Another example of a problem where a quantum computer is much faster than a conventional computer is called factoring: finding a set of numbers that when multiplied together produce the original number given to you. For large enough numbers, it is impossible for conventional computers to solve, and is used to “hide” data in many web encryption protocols. A large enough quantum computer would be able break this encryption.   The issue right now is that, so far, all of the exciting applications of quantum computing seem to require a very big and very complicated quantum computer that’s much more complex than we can build today. While researchers are both working on building better hardware and creating new applications, we still don’t know whether or when there will be broader commercial applications of quantum computing, and what those will be.   Even if quantum computing is very successful, I think most of the computing we do will still be conventional computing, because it’s much cheaper and much faster for many operations. Quantum computers will be used to run the special applications where it provides its performance advantage.

How much progress has been made in quantum computing since your report came out? Recently, the New York Times reported that Google now has a quantum computer that can do tasks that regular supercomputers couldn’t do in 10 septillion years.

Horowitz: When our report came out, it said that there are a couple things that are important to track, in terms of the progress of quantum computing. One is the number of qubits that people can build, and another is the accuracy and fidelity — how many operations can be done before making an error — of the qubit operations. We should also track the algorithms and the applications of these machines.   There’s been substantial progress in actually building the quantum computers we have now. Previously, we had quantum computers with 30 or 40 qubits. Now we have machines with good fidelity and over 100 qubits. For example, I believe the Google system has slightly over 100 qubits.   Also, the best fidelity when the report was released was 99.5%, or roughly an error every 200 operations. Now we have some systems that have 99.9% fidelity, which is five times better. We can now do 1,000 operations before an error. Since errors will occur in quantum computers, error correction will be required in these machines, and there have been a lot of recent advances in error correction — dramatically reducing the number of extra qubits needed to correct errors. But we still need better fidelity to make error correction practical.   A number of new technologies have come along since the report’s release. Recently, there’s been a lot of work in cold or neutral atom quantum computing, and there’s a major effort to build a photonic quantum computer. There has been significant progress in building quantum computers since the report was issued.

Why is it so hard to build quantum computers? What are the barriers?

Horowitz: I think the best way to talk about what the barriers are is to go back to my pool ball analogy. If you have a coin, and it’s sitting on a table, it’s going to stay there. And if you have an operation that flips it, even if you make that operation a little bit different than it should be, it will still be flipped or not flipped, right? It’s kind of robust.   But if you have a ball on the table, and your operations are rotating the ball, first of all, the table had better be pretty still and extremely level. So, you have to be very careful that the environment around the ball doesn’t move it around.   So, to get that kind of stability, with all of these machines, the basic qubits have to be very cold. Now, in some systems, it literally means they’re very cold, but in other systems, it means that they’re highly isolated and cooled by light — so there’s very little stray energy.   And then the second problem is that the operations you do on these have to be extremely precise, because you’re rotating the state a certain amount, and you have to rotate it exactly the right amount; if you don’t, you have a little residual error in your system.

Are there risks linked to quantum computing that policymakers and society should be thinking about?

Horowitz: I think the biggest risk quantum computing poses is that it could perhaps be used to break the security of all encryption that’s used on the web today, as I mentioned earlier. And so that’s the reason that National Institute of Standards and Technology had a competition to develop quantum-secure encryption — and selected winning approaches — so that we can encrypt the communication between computers in a way that a quantum computer can’t break it.   We want to deploy these new protocols early enough so by the time quantum computers can actually break today’s encryption, it doesn’t matter that much, because we moved on to a stronger encryption. One of the problems with network internet protocols is that it takes a very long time to get people to stop using a protocol and shift to a new one. So, even if quantum computers are not going to be available for a decade or more, it’s important that we start doing these changes to the security protocols now.

Your report said it was critical for the U.S. to continue its support for quantum computing, and it sounds like the risk to encryption is one of the reasons. Does quantum computing have any other implications for national security?

Horowitz: National security requires continually exploring new technologies to minimize technological surprise — minimizing the risk an adversary is able to create an attack which you didn’t expect. Since quantum computing’s potential is still not known, it seems wise to continue to explore its potential.