Wednesday, October 23, 2019

IBM explains why Google has a quantum flop

Wired reports:
IBM Says Google’s Quantum Leap Was a Quantum Flop ...

Monday, Big Blue’s quantum PhDs said Google’s claim of quantum supremacy was flawed. IBM said Google had essentially rigged the race by not tapping the full power of modern supercomputers. “This threshold has not been met,” IBM’s blog post says. Google declined to comment. ...

Whoever is proved right in the end, claims of quantum supremacy are largely academic for now. ... It's a milestone suggestive of the field’s long-term dream: That quantum computers will unlock new power and profits ...
Wired says "academic" because everyone quoted claims that quantum supremacy will soon be achieved.

But where's the proof?

Nobody believed the Wright brothers could fly until they actually got off the ground. Quantum supremacy was supposed to be a way of showing that quantum computers had gotten off the ground. If those claims are bogus, as IBM now claims to have proved, then no quantum computers have gotten off the ground. That "long-term dream" is pure speculation.

Update: Google is now bragging, as its paper appeared in Nature. I assumed that it was trying to get into either Science or Nature, but I believe that Nature claims that it does not object to releasing preprints. If so, Google could have addressed criticisms after the paper was leaked.

Quanta mag has an article on the Google IBM dispute:
Google stands by their 10,000 year estimate, though several computer experts interviewed for this article said IBM is probably right on that point. “IBM’s claim looks plausible to me,” emailed Scott Aaronson of the University of Texas, Austin. ...

Aaronson — borrowing an analogy from a friend — said the relationship between classical and quantum computers following Google’s announcement is a lot like the relationship in the 1990s between chess champion Garry Kasparov and IBM’s Deep Blue supercomputer. Kasparov could keep up for a bit, but it was clear he was soon going to be hopelessly outstripped by his algorithmic foe.

“Kasparov can make a valiant stand during a ‘transitional era’ that lasts for maybe a year or two,” Aaronson said. “But the fundamentals of the situation are that he’s toast.”
Following his analogy, IBM's Deep Blue did beat Kasparov, but not convincingly. Maybe the computer was lucky. It was really the subsequent advances by others that showed that computers were superior.

So Aaronson seems to be saying that this research does not prove quantum supremacy, but other research will soon prove it.

We shall see.

Meanwhile, let's be clear about what Google did. It made a random number generator out of near-absolute-zero electronic gates in entangled states. Then it made some measurements to get some random values. Then it said that the device could be simulated on a classical computer, but it would take more time. Maybe 10,000 years more, but maybe just a couple of hours more.

That's all. No big deal, and certainly not quantum supremacy.

Update: Scott Aaronson weighs in , and admits that he was a Nature reviewer. He is happy because he is professionally invested in quantum supremacy being proved this way.

But he admits that Google's claim of 10k years is bogus, and that Google does not have any scalable qubits at all. Further more the researchers cooked the circuits so that they would be hard to simulate classically, while being completely useless for actually doing a computation.

To actually compute something useful, Google would need scalable qubits with some fault-tolerance system, and Google is no closer to doing that.

It has long been known that there are quantum systems that are hard to simulate. The only new thing here is that Google says its system is programmable. I am not sure why that is a good thing, as it cannot be programmed to do anything useful.

Update: Aaronson argues:
But I could turn things around and ask you: do you seriously believe at this point that Nature is going to tell the experimenters, “building a QC with 53 qubits is totally fine — but 60? 70? no, that’s too many!”
The flaw in this argument is that they don't really have a QC with 53 qubits. They have a random number generator with 53 components that act enuf like qubits to generate random numbers.

Computing something useful is expected to require 10 million real (scalable) qubits. Yes, Nature may very well say you can have a quantum device to generate random numbers, but not get a quantum computational advantage.

2 comments:

  1. As far as I can tell, it is an impressive achievement, whether you call it quantum supremacy or not. But I still have serious doubts as to whether this exponential trend will continue and it can be used to factor integers.

    I am also disappointed that there doesn't seem to be a way to directly test it for a small number of qubits other than the way they tested it.

    ReplyDelete
    Replies
    1. If this is impressive, I have a bridge I'd like to sell you.

      Delete