This is the final installment of a four-part series. Read part one here, part two here and part three here.
Jonathan Tennenbaum: Coming back to the Chinese results, to what extent do you consider that they have actually demonstrated “quantum supremacy” over classical computers?
Scott Aaronson: Here is something I would stress: The Chinese system is a fully characterized system. So the Chinese team can and hopefully will publish a complete description of everything they did.
And then they can say to any classical skeptic, OK, then you try to reproduce the results with your classical computer. Here is all the data that you would need to do that. You don’t need to know all the complicated details of this system.
You just need to know some parameters that they publish on the Internet and then you can try with your classical computer to do the same thing.
And the hope is that you will have a hard time doing it and that the reason that you’ll have a hard time is not any complicated details of the experiment. But it is really the exponentiality that is inherent in quantum mechanics.
I would say if that is really the case, then it is fair to call this quantum supremacy.
JT: And what about applications? I understand that there could be a relatively direct application of the Chinese sort of system to calculating certain spectra in chemistry.
SA: Let me put it this way: It is not ruled out. Another potential application is something I came up with a couple of years ago, which is generating cryptographically certified random bits, which you could then use for various applications including cryptocurrency.
Basically, once you have quantum supremacy based on a sampling problem, you can almost immediately repurpose it for that application. Whether or not it will have a market, I don’t know.
JT: I know a person who does very big calculations on the energy states of atomic nuclei, which is a very important area. And you get quickly to the apparent limits of existing supercomputers.
Of course, some mathematician or physicist could come along and make the computation much easier. But my friend will ask: Am I going to live to see the quantum computer that will do this? Should I put my savings into the stocks?
SA: Look, I am optimistic that within the next decade we could start to see the first applications of small-scale quantum computers that are actually useful. I think that by far the best shot is for quantum simulation problems. I know that still doesn’t mean that we’ll have a universal quantum simulation.
JT: Would calculating the energy levels of a nucleus be a quantum simulation in that sense?
SA: Yes. That is the kind of thing you could do with a quantum computer. The hard part is that the classical people have gotten really, really good at developing shortcuts.
So the hard part, if you’re a quantum computing person, is that it’s not enough to do something. You have to beat the best that anyone could do with their classical computer.
JT: So you’re not just competing with classical computers. You are competing with the mathematicians, physicists and computer specialists who are finding ways to simplify the calculations and do them faster.
SA: Classical algorithms are a moving target. What are their limits? This is one of the greatest unsolved problems. That is part of what makes quantum supremacy so tricky.
We see that as soon as someone comes out with a claim of quantum supremacy, people will try to shoot it down by improving how they simulate that with their classical computers.
This happened with the Google thing and that’s happening with the Chinese thing right now.
By the way, we talked about some of the disadvantages of boson sampling – that it’s not universal, at least for now, and not programable. It might have a few limited applications. But I don’t think boson sampling is going to be good for simulating nuclei. There’s just so much interaction in a nucleus. I think you probably want a universal quantum computer for that.
But there is also one advantage of the Chinese experiment over the Google experiment. They have a much larger number of possible outcomes, a much larger state space.
One of the attacks that was proposed against Google’s experiment was to say: If you use a gargantuan enough supercomputer, like IBM’s Summit computer, which has petabytes of hard disk, then you could just explicitly calculate the entire distribution, all nine quadrillion probabilities using two basketball courts’ worth of hard disk.
And then once you’ve done that, you could use it to take as many samples as you want. Then we would take maybe a few days to do that computation. They didn’t actually do it, it is not so easy to get time on the world’s biggest supercomputer. But it is just within the limit of being able to do.
If the Chinese experiment achieved quantum supremacy at all, then it would be beyond that limit. Even the biggest supercomputer on Earth would not have nearly enough disk space to write down the whole distribution.
This is the last installment of a four-part series.