Quantum Supremacy: Made in China?!

If that's all this is, then Google have already done it, isn't it? I think there was a thread about that one.
 
Chanakya said:
If that's all this is, then Google have already done it, isn't it? I think there was a thread about that one.
Click to expand...

There is still some scientific debate about that based on the fact that you probably could do such a calculation on a classical computer.
 
Darat said:
There is still some scientific debate about that based on the fact that you probably could do such a calculation on a classical computer.
Click to expand...

You can do all of them on classical computers. Quantum computers can't do more than classical computers, they can just do some things faster.
 
caveman1917 said:
You can do all of them on classical computers. Quantum computers can't do more than classical computers, they can just do some things faster.
Click to expand...


But that article says, "...In this way, Pan and colleagues could find solutions to the boson-sampling problem in 200 seconds. They estimate these would take 2.5 billion years to calculate on China’s TaihuLight supercomputer ...".

So not really. Not in any meaningful sense. (Unless they're wrong or something? They might be, I wouldn't be able to evaluate that. And besides, the article does say that the specific nature of the problem addressed kind of lends itself to this kind of computing, or something like that, so maybe that could be a factor too, against generalizing from what they're saying there.)
 
Chanakya said:
But that article says, "...In this way, Pan and colleagues could find solutions to the boson-sampling problem in 200 seconds. They estimate these would take 2.5 billion years to calculate on China’s TaihuLight supercomputer ...".
Click to expand...

Yes exactly, classical computers can do it too, they just do it slower.

So not really. Not in any meaningful sense.
Click to expand...

Yes really. And not just in a meaningful sense, in a precise one as well, namely that they are at the same level of the Chomsky hierarchy. Meaning there is nothing that a quantum computer can calculate that can not be calculated by a classical computer, it can only do so faster. In particular, you won't be able to solve the Halting problem or any other undecidable problems on a quantum computer.
 
caveman1917 said:
Yes exactly, classical computers can do it too, they just do it slower.



Yes really. And not just in a meaningful sense, in a precise one as well, namely that they are at the same level of the Chomsky hierarchy. Meaning there is nothing that a quantum computer can calculate that can not be calculated by a classical computer, it can only do so faster. In particular, you won't be able to solve the Halting problem or any other undecidable problems on a quantum computer.
Click to expand...


Google was ~ 100 million times faster than the fastest classical computer. This one's (allegedly) ~ 2.5 billion times.

So you're saying the aim, the end-result, of quantum computing should be to perform computations that cannot be replicated by classical computers ever, not even if they went at it till the end of the universe, is that right?

Fair enough, if that's the case. I wasn't aware.

(Or were you simply responding to Darat's comment, and correcting his impression that quantum computing is not doable by classical computers, without really referencing this Chinese claim at all?)

Either way, 2.5b/100m is pretty impressive. If true!
 
Chanakya said:
So you're saying the aim, the end-result, of quantum computing should be to perform computations that cannot be replicated by classical computers ever, not even if they went at it till the end of the universe, is that right?
Click to expand...

How do you even get there? I've said that quantum computers can not perform computations that can not be performed by classical computers.

Or were you simply responding to Darat's comment, and correcting his impression that quantum computing is not doable by classical computers, without really referencing this Chinese claim at all?
Click to expand...

Yes, the Chinese team doesn't really have much to do with it, I'm correcting a common misconception that quantum computers can compute things that classical computers can't.
 
Came across another news article that discusses this, in (just a wee bit) more detail.

To me, this seems to stand out for two reasons: first, because their advance over Google last year is so very pronounced; and second, because this is China that seems to have done this, and may have stolen a march over the rest, at least for the time being.

What I'm surprised about, is why the news isn't inundated -- to the extent such essentially technical news would, especially in Covid times -- with reports and analyses on this. (The Google thing last year was covered far more widely. On the other hand, that was the first such, so I guess in that sense it was different.)
 
Darat said:
Add in a "practically".
Click to expand...


200 seconds as against 2.5 billion years (!) is somewhat stronger than just "practically".

I'm not anywhere close to well enough informed about this, but I guess what might make a difference is if there's some limit to how fast classical computers can go. If there's nothing that stops classical computers from approaching those speeds in another five or ten or twenty years, then okay, this jump, while sure it's impressive, isn't something fundamentally removed from what went before, in practical terms. But if there's some limit to how fast classical computers can go, so that computation speeds of this magnitude are forever beyond the reach of classical computers, then I'd say the difference is, from an applications perspective, way more fundamental.
 
Chanakya said:
200 seconds as against 2.5 billion years (!) is somewhat stronger than just "practically".
Click to expand...

You are confused about what caveman1917 is saying. He's not saying that quantum computers aren't impressive or aren't this or that. He's making a very clear and precise statement that the class of things that are computable by quantum computers is the same as the class of things computable by classical computers. This is a mathematical fact.

That doesn't mean there's no difference between them. The speed at which those computations can be done is certainly important. That's something he also pointed out, if you noticed.

I may have this wrong, but for some class of problems that difference in speed is one of exponential as compared to linear functions. So when you get a sufficiently complex problem, the total time taken for the computer to solve the problem grows extremely large in the classical case as compared to a potentially short time for the quantum case.

And that seems to be exactly the difference between "possible in theory" as compared to "possible in practice", so I think darat's distinction is a valid one.

I'm not anywhere close to well enough informed about this, but I guess what might make a difference is if there's some limit to how fast classical computers can go. If there's nothing that stops classical computers from approaching those speeds in another five or ten or twenty years, then okay, this jump, while sure it's impressive, isn't something fundamentally removed from what went before, in practical terms. But if there's some limit to how fast classical computers can go, so that computation speeds of this magnitude are forever beyond the reach of classical computers, then I'd say the difference is, from an applications perspective, way more fundamental.
Click to expand...

There certainly are limits to how fast classical computers can go. There are limits related to energy (it takes a certain minimum amount of energy to delete one bit, for instance), and size (there is a minimum size that you can build a transistor, given quantum mechanics, which limits the speed that they can do computation). Those limits apply to both current technology (ie. current computers require a certain amount of energy to do computations, and a certain number of transistors/unit area. There is also a theoretical limit to both of those things for any classical computer, and while we haven't reached those limits, we're not that far from them.
 
Roboramma said:
You are confused about what caveman1917 is saying. He's not saying that quantum computers aren't impressive or aren't this or that. He's making a very clear and precise statement that the class of things that are computable by quantum computers is the same as the class of things computable by classical computers. This is a mathematical fact.

That doesn't mean there's no difference between them. The speed at which those computations can be done is certainly important. That's something he also pointed out, if you noticed.

I may have this wrong, but for some class of problems that difference in speed is one of exponential as compared to linear functions. So when you get a sufficiently complex problem, the total time taken for the computer to solve the problem grows extremely large in the classical case as compared to a potentially short time for the quantum case.

And that seems to be exactly the difference between "possible in theory" as compared to "possible in practice", so I think darat's distinction is a valid one.
Click to expand...

Exactly the difference caveman is pointing out is that while these can address problems that could not be solved in many lifetimes they do not address problems that are impossible for conventional computers to solve. The difference between something that is for all practical purposes impossible because it would take millions or more of years to do, vs something that is theoretically impossible.

The difference between them is not insignificant, it is the difference between saying you don't think something can be done because you don't think the engineering is possible vs it violates very well established laws of physics. Reputable scientists have made all kinds of wrong assumptions about engineering and declaring man will never do something that is done a short while later. This is because of their assumptions on the engineering feasibility of it not usually that it violates well established physics.
 
ponderingturtle said:
Exactly the difference caveman is pointing out is that while these can address problems that could not be solved in many lifetimes they do not address problems that are impossible for conventional computers to solve. The difference between something that is for all practical purposes impossible because it would take millions or more of years to do, vs something that is theoretically impossible.
Click to expand...

Exactly.
 
Having read the article, this does not seem to be applicable to a general-purpose computer. IOW, the only problem it can solve is the one it was designed to simulate.

However, Weedbrook points out that as yet, and in contrast to Google’s Sycamore, the Chinese team’s photonic circuit is not programmable, so at this point “it cannot be used for solving practical problems”.
Click to expand...
 
Roboramma said:
You are confused about what caveman1917 is saying. He's not saying that quantum computers aren't impressive or aren't this or that. He's making a very clear and precise statement that the class of things that are computable by quantum computers is the same as the class of things computable by classical computers. This is a mathematical fact.

That doesn't mean there's no difference between them. The speed at which those computations can be done is certainly important. That's something he also pointed out, if you noticed.
Click to expand...


I appreciate caveman1917’s inputs, because I’ve interacted with him in the past, and I know that he knows, very well, what he’s talking about, as far mathematics and physics (which I, TBF, don’t, not really, not at this level); but in this case this trivial point had been clear from the start. I just hadn’t wanted to keep belaboring this very obvious thing beyond a point; until these further comments now, that is.

Sure, quantum computers are hugely greatly enormously faster than regular computers. But enormously faster is all they are. If a problem does not admit of solution at all, then obviously a quantum computer cannot magically solve it; it isn’t magic after all. That much seems straightforward.

When Darat spoke of whether classical computers were able to do something that this new lot can or not, I took him to mean that he was referring to doing it in a “meaningful”, “practical” sense. If your classical computer can do it, but do it only in as long as it might take for the universe to come to an end, or something of that magnitude, then that’s hardly (the classical computer) being capable of doing it in any “meaningful” sense, right? And, as he clarified later on, that is indeed how he did intend his comment.

Given how obvious this much is, I had wondered—and cross-checked with caveman1917, while also clearly mentioning the other, more obvious, explanation—what he was wanting to convey there. After all, even with my layman’s (non-)understanding, two things, two possibilities stood out, that might tone down the …magnitude of this new development, and I therefore asked about both those: first, whether the limit to speeds that classical computers can achieve is something that might easily be surpassed to quantum levels in future; and second, if the specific nature of the problem that this new computer has solved might limit its wider, more general application.


I may have this wrong, but for some class of problems that difference in speed is one of exponential as compared to linear functions. So when you get a sufficiently complex problem, the total time taken for the computer to solve the problem grows extremely large in the classical case as compared to a potentially short time for the quantum case.

And that seems to be exactly the difference between "possible in theory" as compared to "possible in practice", so I think darat's distinction is a valid one.
Click to expand...


That is what occurred to me as well. Except I wasn’t sure that what had occurred to me was actually the case, since I’m no kind of expert on this sort of thing, not even remotely.


There certainly are limits to how fast classical computers can go. There are limits related to energy (it takes a certain minimum amount of energy to delete one bit, for instance), and size (there is a minimum size that you can build a transistor, given quantum mechanics, which limits the speed that they can do computation). Those limits apply to both current technology (ie. current computers require a certain amount of energy to do computations, and a certain number of transistors/unit area. There is also a theoretical limit to both of those things for any classical computer, and while we haven't reached those limits, we're not that far from them.
Click to expand...


Right, thanks for clarifying that!
 
ponderingturtle said:
Exactly the difference caveman is pointing out is that while these can address problems that could not be solved in many lifetimes they do not address problems that are impossible for conventional computers to solve.
Click to expand...


caveman1917 mentioned one such problem. A discussion of what kinds of problems these are, in terms that layfolks might understand, might be an interesting (if somewhat off-topic) side-discussion to this, if those who’re well informed on this want to go down that route. This Halting Problem thing, for example, that he brought up. (Looked it up just now!)


The difference between something that is for all practical purposes impossible because it would take millions or more of years to do, vs something that is theoretically impossible.

The difference between them is not insignificant, it is the difference between saying you don't think something can be done because you don't think the engineering is possible vs it violates very well established laws of physics. Reputable scientists have made all kinds of wrong assumptions about engineering and declaring man will never do something that is done a short while later. This is because of their assumptions on the engineering feasibility of it not usually that it violates well established physics.
Click to expand...


I understand. Although that was my impression too, it’s good to have people like you, and caveman1917, and Roboramma, who clearly know what you’re talking about here, spell this out clearly, so that layfolks like me aren’t left guessing about this (irrespective of whether that uninformed guess happens to accidentally have been the right one). That’s one of the pleasures of discussing things like this in this forum.
 
Puppycow said:
Having read the article, this does not seem to be applicable to a general-purpose computer. IOW, the only problem it can solve is the one it was designed to simulate.
Click to expand...


Right. That’s what I was wondering, when I read that article. So then, although faster, this Chinese product/project is clearly more limited, in that sense, than Google last year. (Or was the Google development also limited in this sense, limited to only some specific problem or classes of problems? Might general applicability at those phenomenal speeds be something that quantum computing has not yet actually ‘solved’?)
 
Chanakya said:
I appreciate caveman1917’s inputs, because I’ve interacted with him in the past, and I know that he knows, very well, what he’s talking about, as far mathematics and physics (which I, TBF, don’t, not really, not at this level); but in this case this trivial point had been clear from the start. I just hadn’t wanted to keep belaboring this very obvious thing beyond a point; until these further comments now, that is.

Sure, quantum computers are hugely greatly enormously faster than regular computers. But enormously faster is all they are. If a problem does not admit of solution at all, then obviously a quantum computer cannot magically solve it; it isn’t magic after all. That much seems straightforward.
Click to expand...

This point is neither straightforward nor trivial. Hyper-Turing machines could, conceivably, solve the Halting problem. Quantum computers can't do so, however. It's not that computational classes don't exist that can solve the Halting problem, it's that quantum computers are in the same class as classical computers.
 
caveman1917 said:
You can do all of them on classical computers. Quantum computers can't do more than classical computers, they can just do some things faster.
Click to expand...

excepting maybe for true (vs. pseudo)random number generation
 
wea said:
excepting maybe for true (vs. pseudo)random number generation
Click to expand...

Computers now have true random number generation, they use small thermal variations in cpu temperature etc for it. In linux these get written to a special abstract file, and you can make a certain OS call to get some bytes of random data from that file. In windows there is something similar. This is usually used for encryption rather than random number generation, the rate at which this random data gets written is too slow for most rng purposes, but modern classical computers do have true random number generation.
 
caveman1917 said:
This point is neither straightforward nor trivial. Hyper-Turing machines could, conceivably, solve the Halting problem. Quantum computers can't do so, however. It's not that computational classes don't exist that can solve the Halting problem, it's that quantum computers are in the same class as classical computers.
Click to expand...


Oh, OK. I hadn't been aware of this, this hyper-turing machine thing. Thanks for bringing this up!


-------


This is fascinating reading! For those who, like me, might not have known about this, here are some links (from among a much larger number that I speed-browsed through, via Google), that you might find (very!) interesting:

Hypercomputation: Hype or Computation?

Super-Turing or Non-Turing? Extending the Concept of Computation

No Super Turing Machines?

Although this one's a discussion, not a paper or article, but you can click your way along with what they're saying, eavesdrop on their conversation, and in the process get to learn a bit about all of this.

A fifty-pager PDF. I've gone through some of it, and intend to finish it at leisure later on. Although this isn't exactly easy going for the layperson, but once you start reading it through moving lips, much of it turns out to be not entirely as unintelligible as it first appears!


-------


This thread's about quantum computing, and one particular quantum computing endeavor at that, but what the hell.

Fascinating though these hypercomputers / hyper-turing machines are, at this time, as far as I can see, still strictly in the might-be-possible category, involving spacetimes that may or may not actually exist in reality. Of course, I suppose quantum computers also had been just that, not long ago, and now here are people actually developing them, so who knows?

caveman1917, Roboramma, ponderingturtle, Puppycow, and/or anyone else who might be well informed about this, it would be great to have any more inputs you might want to share on this. Including whether these hyper-turing machines might be anything more than theoretical possibilities/formulations at this stage, that is, do we know with any degree of certainty at all whether we might one day be able to actually construct these?
 
Last edited:
Chanakya said:
caveman1917, Roboramma, ponderingturtle, Puppycow, and/or anyone else who might be well informed about this, it would be great to have any more inputs you might want to share on this. Including whether these hyper-turing machines might be anything more than theoretical possibilities/formulations at this stage, that is, do we know with any degree of certainty at all whether we might one day be able to actually construct these?
Click to expand...

Most likely not. But, it should be pointed out here, that we will never be able to construct Turing machines either. A Turing machine requires infinite memory, all we are able to construct in practice are, strictly speaking, finite state machines (the lowest level on the Chomsky hierarchy). When discussing computational classes we're talking about abstract idealized machines.
 
You must log in or register to reply here.

Back
Top Bottom