Quantum Supremacy? Google Claims 70-Qubit Quantum Supercomputer

Google says it would take the world's leading supercomputer more than 47 years to match the calculation speed of its newest quantum computer, reports the Telegraph: Four years ago, Google claimed to be the first company to achieve "quantum supremacy" — a milestone point at which quantum computers surpass existing machines. This was challenged at the time by rivals, which argued that Google was exaggerating the difference between its machine and traditional supercomputers. The company's new paper — Phase Transition in Random Circuit Sampling — published on the open access science website ArXiv, demonstrates a more powerful device that aims to end the debate.

While [Google's] 2019 machine had 53 qubits, the building blocks of quantum computers, the next generation device has 70. Adding more qubits improves a quantum computer's power exponentially, meaning the new machine is 241 million times more powerful than the 2019 machine...

Steve Brierley, the chief executive of Cambridge-based quantum company Riverlane, said: "This is a major milestone. The squabbling about whether we had reached, or indeed could reach, quantum supremacy is now resolved."
Thanks to long-time Slashdot reader schwit1 for sharing the article.

Fake

[backslashdot]

So far, all the "quantum supremacy" claims .. upon reading the actual publication turns out to be using the quantum computer to simulate quantum effects. That's literally like saying you have a ultraviolet light based computer because you shone a UV light on a fluorescent material and it computed what color should be the result faster than a computer could calculate the energy bands. Call me when they can factor prime numbers faster than a computer. THAT would be quantum supremacy.

Re:

[backslashdot]

* factorize composite numbers LOL

Re:

[KiloByte]

On the other hand, if they could indeed factorize prime numbers it'd be a far more profound result than factorizing composite. And only slightly less believable than claims of this quantum machine.

Re:

[itsme1234]

No, there isn't anything "far more profound" in that. Also testing for primality is pretty easy and efficient so no, there isn't anything more there, unless you meant if they could prove prime numbers are composite - which is by definition not the case. This isn't like finding a physical object it shouldn't exist based on some theories that we refine all the time, these are colliding by definition.

Re:

[presearch]

Buncha' Byte Supremacists at Google

Re:

[dvice]

"Previously, the fastest quantum algorithm for this process was Shor’s Algorithm, which required millions of qubits to work; the new algorithm needs just hundreds. "
https://www.azoquantum.com/Art... [azoquantum.com]

So we are not yet there.

Re:

[Entrope]

It is far from clear that the novel claims in that Chinese study are true. https://quantumcomputingreport... [quantumcom...report.com]

Re:

[tlhIngan]

Well, at 70 qubits, it can factor a 70 bit number, theoretically. (Granted, Shor's algorithm requires more, but in theory...).

Given our RSA keys are around 4096 bits, you would in theory need at least a 4096 bit quantum computer.

Note that most of the world has moved on - RSA with its large key sizes requires a lot of computation, so using something like elliptic curves is much faster. Plus, elliptic curves aren't vulnerable to factorization attacks.

Also, likewise, AES is not vulnerable to factorization atta

Re:

[ceoyoyo]

Simulating quantum systems are what we should be doing with quantum computers. Reading people's mail is a stupid human affectation. Demonstrating quantum supremacy by simulating a helium atom would be very impressive.

The claims of quantum supremacy to this point have been mostly sampling random distributions. That can be useful, but it's not exactly an impressive benchmark.

Lame article

[msauve]

What a worthless article. It throws around figures without any reference point. "[T]ake rivals 47 years"? OK, but how long does it take this quantum computer to do the same thing? Then the article repeats the flaw, "it would take Frontier, the worldâ(TM)s leading supercomputer, 6.18 seconds to match a calculation from Google's 53-qubit computer from 2019." Three significant digits of precision vs. none. Although it may not much matter, since "The rival machines were measured on a randomisation task that critics say favour quantum computers and lack any practical value beyond academic study."

I wanted to check the referenced paper, but what should be an obvious link to it ("A paper from researchers at Google") points to some Telegraph article on financial markets.

Re:

[dvice]

I think they didn't mention the time for quantum computer because the results are instant. The idea is that it tries all possible combinations at the same time, so result is found after that "one" try. I think it takes much more time to do the setup for calculations that do the actual calculation.

Re:

[msauve]

Define "instant." Is that 0, or Planck time, or nanoseconds (I/O, at least, would be limited to lightspeed), or blink of an eye, or ???

Re:

[mrthoughtful]

Haha.. I think it's going to be instant as in dessert, not in terms of quantum thermodynamics.

Re:

[msauve]

I'd guess on the order of 6 seconds. Else why would the earlier claim of supremacy over a 6.18 second classical computer have been met with suspicion?

That's not how it works.

[forgotten_my_nick]

> The idea is that it tries all possible combinations at the same time, so result is found after that "one" try.

That isn't how it works at all.

It actually runs 1000's of times and tracks the results. The ones with the most hits are your results.

Think of it of rolling a dice 1000's of times. Eventually all numbers will have a 1 in 6 probability. If the dice is weighted you might find that the number two has a 3/6 chance of appearing.

Re:

[ceoyoyo]

The results are not instant.

One of the big problems with quantum computing is that it's hard to maintain coherence long enough to do an interesting calculation.

Re:

[Shaitan]

"6.18 seconds to match a calculation from Google's 53-qubit computer from 2019"

I think you'll find the devil is in WHAT calculation.

so what is the largest number that it has factored

[KirbyCombat]

I'm just guessing that it hasn't....

a factor of 241 million? I doubt it.

[tal_mud]

The article says: "Adding more qubits improves a quantum computer's power exponentially, meaning the new machine is 241 million times more powerful than the 2019 machine"

They added 17 qubits. That should increase the power by a factor of 2^17 = 131,072 (which is indeed exponential). Instead they used the calculation e^17 = 24,154,952. They read the first three digits as 241 but then were off by a factor of 10. So they had two errors and came up with a number which is more than a factor of 1,000 too large.

Re:

[ERJ]

Your calculations are based on traditional bits (state 0 or 1). Qubits have the equivalent of 4 states (00, 01, 10, 11 in bit terms). So the calculation should be 4 ^ 17 = ~17B.

Doing the math it appears that the scaling number they used was ~3.12 / qubit. I'm sure that someone who knows more about quantum computing could give me a good reason why but my speculation would be that it is maybe because there is a certain error rate inherent in quantum computing so that has to be taken into account.

Re:

[tal_mud]

No, a qubit has only two measurable states (though theoretically it can have a continuum of different mixing values between those two states).

Re:

[jmdevince]

Wrong. https://en.wikipedia.org/wiki/... [wikipedia.org]

Re:

[tal_mud]

They still need two qubits: "...which share a pair of maximally entangled qubits...". Because the qubits are entangled they can *transmit* two bits of information by *transmitting* one qubit. But they still need the two qubits in the first place.

Re:

[Toad-san]

And, despite the math errors, the new machine is accomplishing exactly what? Nada, zip, zilch. I haven't heard of a single thing related to quantum computers' capabilities that seem at all useful ... except maybe the unproven threats to encryption.

Re:

[ceoyoyo]

It's a meaningless calculation.

e^(0.01*17) is 1.2 and is also exponential. So is 0.1^17.

"Google says"

[Opportunist]

Can you prove it?
No?
Then why the hell did you wake me up?

The hype continues

[RUs1729]

And, as of July 2023, existing quantum computers have yet to be able to solve any problems of relevance that cannot be solved by a conventional computer faster and much more economically.

Re:

[timeOday]

"Quantum supremacy" is a term that is bound to draw fire. It implies quantum is better at everything, not just something. Would anybody say that machines achieved "supremacy" over humans in 1644 when Pascal built the first widely-used adding machine? After all it could add at super-human speed. And, heck, addition is really useful.

Re:

[Shaitan]

Indeed.

Specialized problems

[ZectronPositron]

Although many here seem really excited about factoring numbers, here is a list of the actual computations that have been performed with these particular (Googleâ(TM)s) quantum computers: https://quantumai.google/resea... [quantumai.google] One of the people working there told me that theyâ(TM)ve opened up the system for use by researchers (who I presume are essentially buying time or âoecollaboratingâ to use the quantum computers), and letting the researchers identify where this type of computation is act

Re: Specialized problems

[ZectronPositron]

Here is the actual paper - NOT peer reviewed yet https://arxiv.org/pdf/2304.111... [arxiv.org] They mention that theyâ(TM)ve chosen a computation specifically to show quantum supremacy. Also Thai summary article is much better than the Telegraph article https://www.sciencealert.com/g... [sciencealert.com]

The usual bullshit

[gweihir]

First, you cannot actually do anything useful with 70 qbits. Second, the whole supremacy claim is the usual lie by misdirection: The device simulates itself vs. a digital computer simulates it. Yes, for this one irrelevant and useless task, the QC is indeed faster. But not for anything else.

Definition of Quantum Supremacy

[Gaglia]

Before everyone jumps to conclusions, I'd like to point out that the commonly accepted scientific definition of "quantum supremacy" is not what you might think it is. I explained [gagliardoni.net] this back in 2021 already, here is an abstract:

[...] the concept of "quantum supremacy" is a bit underwhelming. You only need to prove that exists: *one* instance size for *one* specific problem that has *no* practical use (and that is specialized for *your* device) in order to claim "supremacy" for your device. And [Richard Borcherds] is totally right on this (but, as I just wrote, he went too far with his argument and did not account for programmability, which makes all the difference.) To which I answer with two observations.

The first is that, for quantum "skeptics" in the community (such as Gil Kalai for example), even this lame definition of quantum supremacy should not be attainable! Skeptics have to understand that denying the feasibility of quantum supremacy is a very difficult task, exactly because the definition is so underwhelming! If you are a non-believer in quantum supremacy, then not only you claim that quantum computers will never have an edge over classical computers for practical problems, but you are claiming that, no matter what the problem and instance size at hand, classical will always be better or roughly on par with quantum in performance. So, think twice before taking this stance. I am looking at you, uneducated Dunning-Kruger commenter on Slashdot.

My second observation is that in my opinion all these misunderstandings are due to the unfortunate choice of the word "quantum supremacy", because it gives false expectations to the non-expert audience.

So, in other words, the definition of "quantum supremacy" is strong, but not as silly as you might expect, it's not "able to simulate itself" but something more. 70 qubits will not allow you to break RSA, but you can still do probably-not-useful computational tasks that are likely out of reach for any non-quantum computer. "Useful" tasks will begin in the realm of chemistry and material science, from what I've heard from colleagues at around 100-110 non-corrected (physical) qubits.