The Chinese Academy of Sciences uses ordinary computers to catch up with the superiority of quantum, and get the calculation of 10,000 years in a few hours

Hengyufa from Quefei Temple quantum | Public account QBITAI

Google quantum computing has claimed that it can complete the task completed in 10,000 years within 200 seconds.

Unexpectedly, this "quantum superiority" is now broken by researchers at the Institute of theory of the Chinese Academy of Sciences:

They used ordinary CPUs to complete the calculation of 10,000 years of calculation within a few hours, and believe that the current quantum computer can be surpassed.

ACM Calculation Prize winner Scott Aaronson also agrees with this conclusion:

I think they are right. If there is a strong enough super computer, they can simulate this task in a few seconds.

Recently, Science has published articles about this, let's take a look at what happened.

Can't surpass? Optimize to show you

Zhang Pan's team of the Institute of theoretical Physics of the Chinese Academy of Sciences published a paper called "Solution of Sycamore's Sampling of Quantum Over -Sampling", showing the entire process of defeating Google quantum computers Sycamore.

For Google's "quantum superiority" task, a new simulation method is proposed in the article, namely the classic algorithm network method.

To say how good it is, you can use many GPU parallel computing.

Using the calculation cluster of 512 GPUs, the team calculated 1 million output mode in 9 trillion digital string (ie, 53 quantum bits and 20 -cycle quantum line sampling), this calculation process took 15 hours.

If you use this algorithm on a super computer, the same task can be completed in tens of seconds.

If this task is done for Sycamore, it takes about 200 seconds; it takes 10,000 years to complete the supercomputer (Google said so).

△ Google quantum computing chip, just a stamp size

The team also adheres to a point of view, just like Sycamore, without having to pursue excessive accuracy.

The team uses a tensor to represent the relationship between the quantum ratio, and describes the interaction of Sycamore quantum bites as a 3D tensor network.

Therefore, they "drilled holes" on the 3D volume network, cut some connections to simplify relationships, reduce calculation complexity, and use accuracy to deduct speed. This method of shrinking the network makes simulation quantum chips possible.

The data proves that during the completion of the task, only 8 lines of code were lost, and the calculation speed increased by 256 times.

And the preservation is kept at 0.37%, higher than 0.2%higher than Sycamore.

In front of various data, the "quantum superiority" was suddenly not so fragrant.

"Quantum superiority" platform, set up

Faced with this result, Sergio Boixo, a chief researcher at Google quantum AI, responded that this was just a short -term overtaking on the track, and the advantage would not last long.

When the era of quantum calculation was just opened, Google did not deny that the classic algorithm would continue to improve and improve this.

However, Google also said:

We believe that this classic algorithm cannot keep up with quantum circuits after 2022.

After all, when Sycamore was born, supercomputers could only be discouraged about the tasks it executed.

Creating quantum computers out is the task of performing those computing volume "drowning" ordinary computers.

For quantum computers, potential solutions for all tasks can be regarded as quantum waves shaking between quantum bits at the same time.

These quantum waves interfere with each other, the wrong answer will offset each other, and the correct output will "bounce".

In other words, this interference allows quantum computers to split a large amount of data to be calculated.

Take Sycamore, for example, it sets any combination of quantum bit to 0, 1 or 0, and 1, which allows it to process a large amount of input at the same time.

Sycamore is a 53 quantum -bit quantum computer. Its quantum circuit is a miniature resonant circuit made of superconducting metals. It can encode any number from 0 to 253 (about 9 trillion), or even encodes all numbers at one time at a time Essence

Google said Sycamore can defeat the supercomputer Summit developed by IBM. Summit has 9216 CPUs and 27648 GPUs.

IBM immediately refuted that Google's so -called the same task super computer needs "1000 years", but it is only on a specific algorithm.

Once the algorithm is optimized, it only takes about 2.5 days, and Summit will make the same task.

Regarding the "quantum superiority", the debate in the industry and academia never stopped.

Zhang Pan's team can use examples to prove that "quantum superiority" can still be defeated by ordinary computers.

One more thing

The research results of the Zhang Pan team highlight the defects of quantum computers compared to traditional computers, which put forward higher requirements for the more leading "quantum superiority" experiments.

Zhang Pan proposed that some practical applications that can be landed should be found to show the advantages of quantum circuits.

Of course, the defense of "quantum superiority" of ordinary computers does not mean the formula of quantum circuits.

Even compared with supercomputers, Sycamore's operation steps and required power are much less.Even Zhang Pan himself said that if Sycamore's preservation is higher, the team's simulation will not be comparable to it ~ Reference article: [1] https://www.science.org/content/ordinary- computers-can-Beat-google-s -quantum-computer-Ater-all? s = 31 [2] https://www.science.org/doi/10.1126/science.365.6460.1364 [3] https://arxiv.org.org/pdf/2111.03011.pdf

- END -