IBM casts doubt on Google's claims of quantum supremacy

Post by sanxiyn » Wed, Oct 23, 2019, 1:10 PM UTC

nevi-me wrote: post #10

I'm a layman on this, but reading IBM's rebuttal, it seems that they're saying "you can't say that a conventional supercomputer would take 10'000 years because you haven't accounted for spilling to disk and other optimisations". I understand that some of these optimisations might be on the processing logic, but doesn't this end up being like comparing apples to oranges? I understand that there's also what seems to be…

Post by AlexCoventry » Wed, Oct 23, 2019, 1:14 PM UTC

akjssdk wrote: post #8

The main takeaway here is figure 1 in this article: they show that for an increasing circuit depth, computation time (on a classical computer) scales linearly. Google claims, on the other hand, that a classical calculation would scale exponentially. This is the basis for the graph in the Google blog [1], which seems to suggest that the Quantum computer can easily reach points (such as qbits=50, cycles=25) which the c…

Post by lallysingh » Wed, Oct 23, 2019, 1:16 PM UTC

AlexCoventry wrote: post #33

akjssdk wrote: post #8

The main takeaway here is figure 1 in this article: they show that for an increasing circuit depth, computation time (on a classical computer) scales linearly. Google claims, on the other hand, that a classical calculation would scale exponentially. This is the basis for the graph in the Google blog [1], which seems to suggest that the Quantum computer can easily reach points (such as qbits=50, cycles=25) which the c…

So if they increase the number number of qubits they'll be have a classically infeasible calculation again? Should be interesting.

Post by andrewla » Wed, Oct 23, 2019, 1:18 PM UTC

Post by AlexCoventry » Wed, Oct 23, 2019, 1:20 PM UTC

htfu wrote: post #22

excalibur wrote: post #7

Earlier quoted context omitted.

But it's still pretty fuzzy. If we agree that 10,000 years on a classical system is supremacy but 2.5 years isn't, where's the cutoff? 25 years? 250? An average human lifespan? Edit: Thanks guys for pointing out where I misread the article, leaving mistake intact so the replies make sense.

2.5 _days_, not years. A still fuzzy and arbitrary but imo somewhat elegant cutoff would let feasibility be set by the market and technological development, in the sense of something not being worth throwing compute at _now_ because the timescale of work is slow enough that future classical efficiencies will progress (much) quicker.