Jonathan Tennenbaum: Here’s a question that the readers of Asia Times will be especially interested in. How would you compare the effort in China in the area of quantum computing with that in other parts of the world, for example, in the United States? Are the Chinese way out ahead? You hear they’re putting a lot of money into this field.
Scott Aaronson: Well, I would say that, a few years ago, China took a clear lead in the area of quantum communication. Communication is different from quantum computing. They launched a satellite mission that can distribute photon entanglement over 3,000 miles from one end of China to the other end of China. And they can do quantum cryptography over that kind of distance. At least if the weather permits.
It’s not practical yet, but they were the first to demonstrate that you can send a qubit to space, have it come back to Earth and still measure it as a qubit. The US considered doing that and decided not to fund it. That was more than a decade ago.
JT: Interestingly, it was exactly the same Chinese research group around Pan Jianwei, which realized the satellite quantum communication experiment, that also built the Jiuzhang,
SA: That is what helped spur all of this panic in the US, that there is a “quantum Sputnik.” That helped lead to the passage of the National Quantum Initiative Act, which Congress unanimously passed a couple of years ago. Imagine, what does Congress still do unanimously?
JT: Yeah, a rare coherence in Washington these days.
SA: Actually, the Senate sponsor of the bill was Kamala Harris. I met with her staff when they were writing it. It’s a billion dollars or so for quantum computing research. People in Congress noticed that China was pulling ahead in photonic quantum communication, and in their minds I think quantum communication and quantum computing were all kind of conflated.
JT: Somehow just the use of the word “quantum” tends to put people into a state of awe.
SA: Yeah, it is an amazing word.
But I would say that the US has had a pretty clear lead in quantum computing per se, in terms of what we know about.
In superconducting qubits, the biggest competitors have been Google’s group in Santa Barbara and IBM’s in Yorktown Heights. There is a startup company called Rigetti in Berkeley. And trapped ion qubits at the University of Maryland and at Honeywell in Colorado.
Concerning photonic qubits, there is a fairly secretive startup company called PsiQuantum in Palo Alto. They are trying to build a photonic quantum computer. But in contrast to the Chinese group, they said: We don’t even care about quantum supremacy. We’re just going to skip right over it. We want to go straight for a full error-corrected quantum computer. They raised more than two hundred million dollars.
JT: The Chinese, whether they are doing it wittingly or not, are doing pretty good for the stocks of certain American companies, I suppose.
SA: It is fairly new, within the last five years, that Google, IBM, Microsoft, Amazon and so on have all gotten into quantum computing to the extent that they have. Partly it’s for prestige or something like that, and partly they want to think 10 or 20 years into the future.
JT: Also I think people in the US are scared by the thought of China turning out millions of Chinese quantum computer PhDs.
SA: I would say that this announcement does represent an entrance of China into scalable quantum computing in a major way. Of course, China has been doing stuff for a long time, but it wasn’t obvious if they were going to be competing at the scaling frontier. At least in photonic quantum computing, it now seems pretty clear that they are.
They have put a lot of eggs into the basket of a photonic quantum computer. I should be clear that we don’t know yet which of these approaches, if any, is going to be the right way to scale up to thousands or millions of qubits.
I would stress that there are at least five approaches that are being pursued. Superconducting qubits, Google, IBM and Rigetti. Photonic qubits, the Chinese and PsiQuantum. Trapped ion qubits, with IonQ and Honeywell.
Neutral atom qubits with the Mikhail Lukin group at Harvard. And then maybe I would include silicon qubits, with a group in Australia doing that. And then there is topological qubits, which Microsoft is doing, which will require creating a new state of matter that’s never been seen in nature.
JT: So you might actually end up with a whole array of different technologies that would be usable for quantum computing.
SA: Right. One thing you can imagine is that you would have different kinds of qubits that are good for different purposes. Quantum computers may even involve hybrids of several of these technologies.
Now if you look at the history of classical computing, another thing you could imagine is that someone will invent the analogue of the transistor. Something so good that it will destroy everything else. I think it’s fair to say that we do not yet have the quantum computing analogue of the transistor. We’re just barely entering the vacuum tube era.
JT: Getting to another question, I can imagine somebody could come along and say: Well, this sounds all great. But what you basically have, if one is not over-awed by the word quantum, is just an analog computer system. I myself have always believed that sooner or later, analog computers will become as important as digital ones. I’ve looked at things like protein folding.
The so-called Levinthal paradox suggests that predicting the conformation that a protein will fold into, from the original chain of amino acids, will generally take far too long with existing computers.
SA: But you know, just a couple of weeks ago, DeepMind said that they’ve solved protein folding. Although it is not clear what “solved” means. But it is clear that there is progress.
JT: What I was getting at is that there could be a certain point where you say: Just let the protein fold. And then I know what the conformation is.
SA: Of course.
JT: So nature solves these problems without computation. Or maybe we can use the protein as a sort of analog computer. So here are some questions. Apart from computational advantage or supremacy, the present digital computers have the great advantage of a kind of universality. They may be slow, but you can program them to do whatever you want.
What about quantum computers? I’ve heard the term “universal quantum computer” but, practically speaking, the systems we are dealing with right now are optimized for specific types of problems.
SA: There are a bunch of things to respond to here. The first thing I would say is that, with the term analog computer, you have to be careful because a quantum computer is in some respects like an analog computer and in other respects like a digital one.
It really doesn’t fit easily into that dichotomy at all, because amplitudes in quantum mechanics are a continuous quantity but as soon as you make a measurement you get a yes or no answer.
The whole theory of quantum error correction, which developed in the ’90s, shows that a quantum computer actually behaves a lot more like a digital computer in some respects. This was very surprising to people.
Concerning the question of universality and programmability, I would say one advantage of Google’s route, over the Chinese one, is that Google’s design actually is universal. Of course, you are severely limited in how many qubits you have and how many operations you can do with those qubits while maintaining the quantum state. But except for those limitations it is universal. You can do an arbitrary sequence of operations, any quantum computation that you want.
And it’s fully programmable. The next second, you can give it a completely different sequence of operations and it will do that one instead.
As for the Chinese experiment, first of all, boson sampling is not a universal model for quantum computing. And we knew that from the very beginning.
JT: It has a different purpose.
SA: Yes, but you can hope anyway that at least your boson sampling experiment would be reconfigurable, meaning that you could change around the beam splitters from one of the experiments to the next. The Chinese experiment is not reconfigurable. They decide once and for all what the pattern of beam splitters should be. They build that out and do the sampling with that.
On the Levinthal paradox: If it is a real paradox rather than just an apparent one, then a folding protein would already count as a computer that seems to work for some task. But you know, it could take decades of work, but I think ultimately there is nothing going on there that a classical computer cannot tell you.
JT: It is a quantum many-body problem.
SA: The one possibility of protein folding being truly hard in the theoretical sense is if quantum effects are actually relevant to how the protein folds – although the brand-new results from DeepMind seem to suggest that you can get answers to within chemical accuracy without needing to account for the quantum effects. That makes sense because proteins are pretty big.
But in some sense it is true that you are building a specialized quantum system that is just trying to solve the specific problem of simulating itself.
The concluding part of the four-part series will be published soon. Check Asia Times regularly for the next installment.