Quantum Supremacy From Google? Not So Fast, Says IBM. (technologyreview.com) 80
IBM is disputing the much-vaunted claim that Google has hit a new milestone. From a report: A month ago, news broke that Google had reportedly achieved "quantum supremacy": it had gotten a quantum computer to run a calculation that would take a classical computer an unfeasibly long time. While the calculation itself -- essentially, a very specific technique for outputting random numbers -- is about as useful as the Wright brothers' 12-second first flight, it would be a milestone of similar significance, marking the dawn of an entirely new era of computing. But in a blog post published this week, IBM disputes Google's claim. The task that Google says might take the world's fastest classical supercomputer 10,000 years can actually, says IBM, be done in just days.
As John Preskill, the CalTech physicist who coined the term "quantum supremacy," wrote in an article for Quanta magazine, Google specifically chose a very narrow task that a quantum computer would be good at and a classical computer is bad at. "This quantum computation has very little structure, which makes it harder for the classical computer to keep up, but also means that the answer is not very informative," he wrote. Google's research paper hasn't been published, but a draft was leaked online last month. In it, researchers say they got a machine with 53 quantum bits, or qubits, to do the calculation in 200 seconds. They also estimated that it would take the world's most powerful supercomputer, the Summit machine at Oak Ridge National Laboratory, 10,000 years to repeat it with equal "fidelity," or the same level of uncertainty as the inherently uncertain quantum system.
As John Preskill, the CalTech physicist who coined the term "quantum supremacy," wrote in an article for Quanta magazine, Google specifically chose a very narrow task that a quantum computer would be good at and a classical computer is bad at. "This quantum computation has very little structure, which makes it harder for the classical computer to keep up, but also means that the answer is not very informative," he wrote. Google's research paper hasn't been published, but a draft was leaked online last month. In it, researchers say they got a machine with 53 quantum bits, or qubits, to do the calculation in 200 seconds. They also estimated that it would take the world's most powerful supercomputer, the Summit machine at Oak Ridge National Laboratory, 10,000 years to repeat it with equal "fidelity," or the same level of uncertainty as the inherently uncertain quantum system.
Three orders of magnitude faster or six? (Score:3)
I think they're kind of missing the point. Whether it's three orders of magnitude faster (like IBM reportedly says) or six (like the original Google paper reportedly says), IMO, the neat thing is that it is possible at all. :-)
Re: (Score:2)
Three additional orders of magnitude is the difference between claiming you lifted a power wheels toy car with your bare hands, or a fully loaded 18 wheeler. It's not like the difference makes no difference.
Re:Three orders of magnitude faster or six? (Score:5, Insightful)
True, but that's only because the act of lifting something with your hands isn't interesting in and of itself. If you told me that you lifted an 18-wheeler with only the power of your mind and it turned out that it was only a toy car, it would still be an interesting story. :-D
Re: (Score:2)
Re: (Score:2)
*Raises eyebrow*
Intriguing.
Re: (Score:1)
That has to be Slashdot's worst car analogy ever. It would be true if the lifting was done with telekinesis however.
Re: (Score:3)
That's the problem: in order to even approach quantum supremacy everyone has to play with toy problems that are stacked in favour of the quantum. As a result, the results are useless, and it takes a *lot* of imagination to extrapolate to where they might be useful. Nobody can even agree on whether the imaginary line has actually been crossed or not.
It is a very interesting question whether a quantum computer can do anything a classical computer can't. In that sentence, "classical computer" doesn't mean a PC
Re: (Score:2)
Theoretically, there are class of problems for which quantum computers have zero advantage and hence they can never be faster than classical computer. E.g. Adding two numbers. So yes, you have to use only those problems where quantum computers are fast (and there are vast category of them) and there is nothing wrong with it.
Quantum computers will never achieve their promise (Score:2)
What they promise is exponential speed up. That means searching a space of N bits in size in N time not 2^N.
At the moment schorrs algorithm is sqrt(n) and another less well known algorithm is cube root. And hey that is fantastic!!! You can turn a 2^15 problem into 2^5? Wow. Not being sarcastic, But it's a gazillion year me far from exponential
But there exists a gulf between theory and experiment due to noise and coupling patterns. I have two ill informed gut feeling. One is that the more bits you coup
Re: (Score:2)
Maybe your gut feeling is based (more or less) on the notion that we do not see quantum effects in the macro world. To see them, significant numbers of qubits or whatever particles with quantum properties you choose must be synchronized in some way. However, quantum jitter prevents this beyond a certain point. I think of these processors as determining that point.
Like you, I'm dubious.
Re: (Score:2)
Actually, the question is scaling. If QCs are only workable for small inputs, then they are useless. For example, there is a very simple factoring algorithm that runs in O(1) if you only have, say, 20 bits or s: Do a pre-computation and then use table-lookup. Interestingly for all input sizes where that approach works, a QC can never beat a classical computer.
It currently looks very much like QCs are only workable for really small inputs or really simple computations. For example, you cannot do factorizatio
Re: (Score:2)
Did you miss the "toy problems" part?
Generating random numbers isn't exactly an interesting problem. Yes, choose things that a quantum computer should be good at, but pick something convincing and potentially useful.
Re: (Score:2)
Here are my objections with more precision.
The task Google chose is generating pseudorandom samples drawn from a 54 dimensional distribution. That distribution is special, in that the different dimensions exhibit quantum interference with each other. The problem is chosen precisely to fit Google's hardware: 54 noisy quantum qubits. Because of the interference, a classical computer has to compute all of the interactions between all of the dimensions in order to get the pattern right.
The quantum computer is c
Re: (Score:2)
It is a very interesting question whether a quantum computer can do anything a classical computer can't. In that sentence, "classical computer" doesn't mean a PC with an x86 processor, it means any computer based on classical computation. To convincing demonstrate that you need to solve a problem that's got some good mathematical reasons why the actual computation is of a certain difficulty.
My take is that with the really bad scaling of QCs, even stacking the deck massively against the classical computer we may not ever see Quantum Supremacy. Of course, if you also get the power of the classical computer massively wrong and far, far too low, then you can give the appearance of having gotten there. But remember that Google is an ad company. They do not care about research, they care about publicity.
Re: (Score:2)
But read very carefully. The claim is not (apparently) that the quantum computer solved the PROBLEM in 200 seconds that would take the largest supercomputer 10,000 years (or 2.5 days) to solve. It's that it solved the problem in 200 seconds that would take the conventional computer 10,000 years (or 2.5 days) to emulate the quantum computer solving the problem.
Further, it was a problem designed to be solved quickly by the quantum computer, kinda like you can design a drag race such that a human runner in ave
Re: (Score:2)
Further, it was a problem designed to be solved quickly by the quantum computer, kinda like you can design a drag race such that a human runner in average condition can beat a top fuel dragster, but the inevitable result doesn't really tell you much.
Thanks, now I had to do this.
Easiest way seems to be that you need to do an unassisted run in each direction, counting the time from when you left one end to when you returned to it. Would also need to stipulate you come to a safe stop at the end.
I wonder at what distance the human is beaten by the dragster.
Re: Three orders of magnitude faster or six? (Score:1)
Re: (Score:2)
Even simpler, the race is a 3 meter sprint. While the dragster is spinning it's wheels madly and inching forward, the human is crossing the finish line. Another second later and the dragster gets traction and disappears down the track at speeds well beyond human legs, but the race is already over by then.
Re: (Score:2)
That is, they said this specific computer [top500.org] would take 10,000 years to solve the same problem.
They say this specific computer would take 10,000 years to complete a run that simulated the quantum computer solving the problem.
They said nothing about how long it would take running a program actually meant for a conventional computer.
That's like saying a 4 banger calculator is faster than a PC simulating the calculator at the subatomic level. True but meaningless.
I'm guessing the 200 seconds doesn't include the time it takes to get the computer configured and loaded in order to run the program.
Re: (Score:2)
"Solved the problem" is a bit of a misnomer too. The quantum computer provided a million samples of noise exhibiting quantum interference. The classical computer had to simulate the quantum mechanics to get the noise pattern right, which meant simulating all the interactions among every pair of qubits. This is something that a digital computer is exceptionally bad at. Could you build an analog (non-quantum) computer that's better? Probably. Could you build one that's as fast as Google's quantum machine? Goo
Re: (Score:2)
But read very carefully. The claim is not (apparently) that the quantum computer solved the PROBLEM in 200 seconds that would take the largest supercomputer 10,000 years (or 2.5 days) to solve. It's that it solved the problem in 200 seconds that would take the conventional computer 10,000 years (or 2.5 days) to emulate the quantum computer solving the problem.
That dishonest nonsense again. Have they not learned anything?
Re: Three orders of magnitude faster or six? (Score:2)
Re: (Score:2)
The claim of quantum superiority by Google, if true, would indicate that no non-quantum system (so nothing that's not super cooled and super isolated) can beat a quantum computer at some tasks.
Re: (Score:2)
I think they're kind of missing the point.
They don't. They point out that a) the problem solved is pretty irrelevant and b) a classical computer still does it in time to be entirely useful. In essence, they point out that (again) Google's "research" is not very good and that the claims made by Google are (again) vastly overstated.
Google to IBM... (Score:2)
Re:Google to IBM... (Score:5, Funny)
Quantum-world uses cat-years.
Excuse me? (Score:3)
Google specifically chose a very narrow task that a quantum computer would be good at and a classical computer is bad at
Well, isn't that the whole point of developing quantum computers?
Factoring huge prime numbers is another example of such a narrow task, and a lot of people are very interested in that particular problem.
Re: (Score:2)
That's a lot of investment to generate some random numbers.
Re: (Score:2)
Was there ever anything new that wasn't considered by some naysayer to be too large of an investment? Isn't that like the theme of the world? Let's not build more houses because investment? Let's not collect food for the winter because investment? Let's not wake up tomorrow because investment?
Re: (Score:2)
Your examples are specious. We invest in collecting food for the winter because we expect a return. If the goal of quantum computing is to generate random numbers, we're throwing away money.
The goal of quantum computing is *not* to generate random numbers. Furthermore, Google claims their demonstration (of generating random numbers) proves something fundamental about the universe. I don't think that claim, or the claim of quantum supremacy, is properly supported by this experiment, as it would be by other p
Re: (Score:2)
Factoring huge prime numbers is not really a challenge for classical computers. I can think of an O(lg n) algorithm that runs on a slightly modified fax machine.
Semiprimes, of course, are significantly harder.
Re: (Score:2)
True enough, I mispoke. Factoring a huge prime number would be rather silly.
Re: (Score:2)
Bill Gates "misspoke" the same way in "Road Ahead" book, so you are not alone!
Re: (Score:2)
I would hope that you could come up with an algorithm that is O(1) for factoring prime numbers no matter what the size of the prime numbers.
Re: (Score:2)
Well, you see, very large primes take more pages to fax. That's where the lg n comes in.
Re: (Score:2)
Too bad that's not the problem Google figured out how to solve.
Re: (Score:1)
I don't get the "Quantum" part?
It was bullshit from day one. (Score:1)
Called it.
Re: (Score:1)
Yes, but our bullshit spews faster!
Re: (Score:2)
To be fair, it was very easy to do so. Not only are most claims about what QCs can do bullshit, but Google's research usually sucks badly in addition.
Quantum IT guys? (Score:2)
Just like in the old days, computers were huge and only scientists used them. I wonder how far we are from having some local quantum IT guys. And do we call them QIT guys?
Re: (Score:2)
Since most homes don't have an adequate supply of cryogenic helium, any end-user facing applications of QC will be done in the cloud.
You can already rent QCs [techcrunch.com] in the cloud, but so far there are no front-end apps.
Re: Quantum IT guys? (Score:1)
Re: (Score:2)
Quantum computing won't follow the same pattern because it requires cryogenic cooling. There's no way around it. And cryogenic cooling requires a certain amount of infrastructure.
The phone in your pocket is a marvel of engineering, but the physics isn't really different from the first electronic computers, barring a few tricks needed when things started getting really small.
Putting a quantum computer in your pocket isn't an engineering problem, it's a physics problem.
Re: Quantum IT guys? (Score:1)
Re: (Score:2)
Where's that guy with binary for his username? He's a troll, but he's got a nice troll that's spot on in this case. The laws of physics disallow some things. It's possible they're wrong, sure, but don't bet on it. Or rather, do bet on it. How much have you got?
Re: (Score:2)
It might well follow the same pattern. Currently it's a little less far along than Charles Babbage. What might that suggest about the time frame?
There's a built in problem (Score:5, Funny)
Re: There's a built in problem (Score:1)
Re: (Score:1)
Schrodinger's cat: "Hey, wow, I don't know how I ended up with such a giant meat cannon, but I'm keeping it, guys!"
Re: (Score:2)
As long as it is hard ...
Re: There's a built in problem (Score:2)
Entangle them first, measure one and infer about the other.
Good luck getting that image out of your heads.
Product Hype (Score:2)
Google simply learned from NVIDIA on how to make your product look much better than it actually is.
The former picks a problem that a Quantum system excels at vs Non-Quantum.
The latter simply writes a " special " driver for the benchmarking folks to use.
Moral of this story: Never believe the hype.
Re: (Score:2)
More likely, they -- and Americans in general, feel threaten by the Chinese who keep eclipsing world records [phys.org], also set by the same Chinese team, in quantum entanglement. So they resort to producing inflated claims like this and 72-bit quantum chip [quantumcom...report.com] without telling us how many bits it can actually entangled together.
We called this one when ... (Score:2)
... it first hit [slashdot.org].
[CaptainDork] I call bullshit.
It's weasel all the way down.
The leading academia do not support this level of technology.
The problem is... (Score:2)
That the fundamentals of quantum physics are flawed.
There is no superposition, sorry. Don't shoot the messenger.
"Superposition" as we know it is the result of a force we are not aware of or are not yet including in the formula.
Re: (Score:2)
There is no superposition
That is my guess as well and I think we will eventually learn that superposition is a combination of:
1) the interference of a real wave phenomenon, like the pattern we observe when sending one particle at a time through a double slit
2) the non-local entanglement of photons where measuring one of the entangled photon seems to instantly collapse its new state to the other photon. And where a particle is something like a photon (maybe entangled with itself) going in circles
But I am no scientist and tha
Re: (Score:2)
1) the interference of a real wave phenomenon, like the pattern we observe when sending one particle at a time through a double slit
This is exactly what I'm talking about. All of the math stems from the assumptions made surrounding the double slit experiment.
The particles are not behaving as a wave, the wave function is not "collapsing" upon observation.
There is a force acting upon the particles that makes them behave as a wave. That is why the math is broken. This "force" is not accounted for.
Re: (Score:2)
That is why the math is broken.
Oh, the math is not broken at all. The problem is rather that this math may not actually apply to reality.
Re: The problem is... (Score:1)
Re: The problem is... (Score:2)
Re: (Score:2)
Well, congratulations, quantum field theory basically says your #1 is correct, with the modification that "real wave phenomenon" is instead "evolution of the wave function."
That doesn't mean superposition doesn't exist, just like it doesn't mean a wave doesn't exist at more than one point at the same time.
Re: The problem is... (Score:2)
Re: (Score:2)
You're correct. QFT doesn't say that waves and particles propagate together. That's a feature of some interpretations of early quantum theory. QFT has quantum fields, which have specific properties, and interact in specific ways. Those fields can be excited, and those excitations can be interpreted as one or more particles or waves, depending on what kind of experiment we use to observe them.
You can think of waves in a quantum field, or particular excitations being particles, but that's just a metaphor, and
Re: (Score:2)
Superposition definitely exists. You can measure it. If you're reasonably resourceful you can probably demonstrate it at home.
I don't think superposition exists in the sense that a photon is in a mixture (superposition) of states until measured. Instead I believe a photon is in a definite state at any given time. When you measure the polarization of a photon at a given orientation of the measurement device you can predict with 100% certainty what state you will measure it at when using the same or perpendicular orientation. This suggests to me that the photon has a specific state and is never actually in a superposition of sta
Re: (Score:2)
Yes, the interpretation of what superposition actually is, is, well, an interpretation.
You're describing a hidden variable theory. There are some pretty hefty restrictions on what kinds of theories are allowed, but hidden variables are one of them. It requires that you accept one of two other implications though:
1) those variables are non-local, which means that influences can travel instantaneously over any distance
or
2) superdeterminism (maybe) which requires something along the lines of a fixed and unchan
Re: (Score:2)
Quantum field theory deals with statistics and unless you think the universe is fundamentally statistical it is not a final theory. And if you piled up a list of issues with QFT besides this one, it would be a long list. So QFT is not a good theory to directly tr
Re: (Score:2)
That the fundamentals of quantum physics are flawed.
There is no superposition, sorry. Don't shoot the messenger.
"Superposition" as we know it is the result of a force we are not aware of or are not yet including in the formula.
With the present demonstrated "quantum" computations, that is entirely possible. This is, IMO, also the only point to continue to do research in this direction: It may lead to some new fundamental Physics because the current standard model may observably _not_ pan out.
Caltech is spelled Caltech - it is NOT intercapped (Score:3)
Yawn (Score:2)
The fact that it wasn't a HUGE story back when Google made the claim, with black helicopters hovering silently above every significant Google employee, was very telling. Either Google was selling a lot of bullshit or the government has had actual, useful quantum computers for some time.
If actual quantum supremacy ever happens (and many think it won't), it would upend the entire world.
None of the above (Score:2)
In the US it will be associated with the NSA and IBM might be involved. Or it could be somewhere in China. Russia also has something running, but they are likely behind the US and China. Israel is also a dark horse possibility.
If you know about it then it is not in the quantum computing top 5. Get used to it.
Hoax? (Score:2)
I just have a feeling that this whole quantum computing thing is a hoax.
Re: (Score:1)
Maybe. Or maybe not. Personally, I'm waiting for demonstration of Shor's or Grover's algorithm on a Quantum computer yielding results that can't be obtained in reasonable time by classic computers. Then we will have genuine quantum supremacy. And I have the uneasy feeling that it won't happen in the next two or three decades, if ever. But miracles happen. Just don't hold your breath.
Quantum Computing Overlords (Score:1)