Google: Demonstrating Quantum Supremacy

[Auto-Didact]

TL;DR

Google has publically claimed to have achieved quantum supremacy using their 53 qubit quantum computer. They have just posted a video on this alongside their publication in Nature.

Here is the paper, published today in Nature:
Arute et al. 2019, Quantum supremacy using a programmable superconducting processor

The promise of quantum computers is that certain computational tasks might be executed exponentially faster on a quantum processor than on a classical processor$^1$. A fundamental challenge is to build a high-fidelity processor capable of running quantum algorithms in an exponentially large computational space. Here we report the use of a processor with programmable superconducting qubits$^{2,3,4,5,6,7}$ to create quantum states on 53 qubits, corresponding to a computational state-space of dimension $2^{53}$ (about $10^{16}$). Measurements from repeated experiments sample the resulting probability distribution, which we verify using classical simulations. Our Sycamore processor takes about 200 seconds to sample one instance of a quantum circuit a million times—our benchmarks currently indicate that the equivalent task for a state-of-the-art classical supercomputer would take approximately 10,000 years. This dramatic increase in speed compared to all known classical algorithms is an experimental realization of quantum supremacy$^{8,9,10,11,12,13,14}$ for this specific computational task, heralding a much-anticipated computing paradigm.

 

[Greg Bernhardt]

IBM is casting a little doubt.
https://www.sciencemag.org/news/2019/10/ibm-casts-doubt-googles-claims-quantum-supremacy

A study from Google claiming quantum supremacy, accidentally leaked online last month, has now been published in Nature. The Google group reiterates its claim that its 53-qubit computer performed, in 200 seconds, an arcane task that would take 10,000 years for Summit, a supercomputer IBM built for the Department of Energy that is currently the world's fastest. But IBM appears to have already rebutted Google's claim. On 21 October, it announced that, by tweaking the way Summit approaches the task, it can do it far faster: in 2.5 days. IBM says the threshold for quantum supremacy—doing something a classical computer can’t—has thus still not been met. The race continues.

 

[jedishrfu]

There are many complexities to quantum computing that must be overcome. The Google result is akin to Edison's first light bulb and now we need the power grid to complete the infrastructure.

As an example, this grad student, Urmila Mahadev is working on whether we can verify QC results:

https://www.quantamagazine.org/graduate-student-solves-quantum-verification-problem-20181008/
and this undergrad student, Ewin Tang in 2018, determined that a QC only algorithm can really be computed just as quickly on a classical computer thus raising the bar for Quantum Supremacy:

https://www.quantamagazine.org/teen...to-quantum-recommendation-algorithm-20180731/
At our university, since 2018, we have a working analog quantum emulator:

https://ieeexplore.ieee.org/document/8638597

 

[f95toli]

IBM is casting a little doubt.
https://www.sciencemag.org/news/2019/10/ibm-casts-doubt-googles-claims-quantum-supremacy

The general feeling in the community is that IBM is being somewhat childish
Also, they are also -somewhat ironically- making somewhat inflated claims about their own paper (which has not been peer-reviewed but was posted on the arXiv on Monday)

 

[nsaspook]

Did IBM demonstrate (in a childish way or not) that the Google problem was not as hard as Google thought it was?

 

[jedishrfu]

This is a classic battle of Titans. IBM is a computer company and this is a turf war. IBM history is full of these battles where in the end IBM wore out the competition.

Also historically, there was a notion that one could never go wrong purchasing an IBM mainframe and so corporate management bought IBM products over other competitors. If and when there was a hardware issue IBM would send in its best engineers to repair it while at the same having a new mainframe ready to ship at a moments notice, at their main site in Poughkeepsie NY. That kind of service was hard to beat.

 

[f95toli]

Did IBM demonstrate (in a childish way or not) that the Google problem was not as hard as Google thought it was?

No, they are claiming that by using a different method and by using more memory (64 peta-byte of hard-drive space); you should be able to do run the same calculation in 2.5 days (as opposed to 10 000 years).
They haven't actually done this; they are extrapolating from existing results on smaller problems.
In order to actually try they would need exclusive access to an upgraded version (much more HD space) of Summit.

Personally, I don't think anyone cares if is 2.5 days or 10 000 years; the Google machine is still at least x1000 faster than the biggest supercomputer on Earth .
Also, publishing the blog-post and the paper 2 days before the paper is published is bad form . Note that IBM have also promised to put a put a 53 qubit machine on their cloud this October. They fact that they picked 53 is quite ridiculous since they don't even have processor of that size, meaning they would probably just use a subset of the qubits of something bigger.
(it is not a secret that the Google's processor has 54 qubits. but one is not working well)