Can’t program, burn money, useless? Pan Jianwei’s live response to the “nine chapters” quantum computing controversy

In 200 seconds, 76 photons passed through the optical network carefully constructed by Pan Jianwei’s team of University of science and technology of China, and Gaussian Bose sampling of 50 million samples was completed. It will take 600 million years for the same math problem to be handed over to Fuyue, the world’s top supercomputer, and the gap is more than one million billion. < / P > < p > this quantum computing model machine, named “nine chapters”, which was unveiled on December 4, is the second quantum computing experiment in the world to reach the “quantum hegemony” standard proposed by Caltech professor preskiel. < / P > < p > Pan Jianwei thinks this name is too overbearing and not academic enough. He chose the word “quantum superiority” with the same meaning that quantum computers outperform the world’s best performing classical computers on specific problems. < / P > < p > despite the choice of a more low-key term, “nine chapters” has caused a sensational discussion and even controversy. Is it just a “useless” question? It’s 10 billion times faster than Google’s “Platanus acerifolia.” how do you figure it out? Is it worth burning money on a machine that can’t be programmed? On December 30, pan Jianwei and Lu Chaoyang, the corresponding authors of the “nine chapters” paper, walked into the X intellectual studio of Mozi salon to have a round table discussion with five other scholars in the field of quantum information or computer. Dispute 1: is “nine chapters” 10 billion times faster than Google’s “Suzuki”? < / P > < p > Google is the first company in the world to complete this challenge. In 2019, Google used 53 superconducting qubits to make a processor called “Suzuki”, which runs random quantum circuit sampling, and takes about 200 seconds to conduct 1 million samples. It will take one year for the best supercomputer, summit computer of Oak Ridge National Laboratory in the United States, to get the same results, and the gap is about 100000 times. < / P > < p > as mentioned above, on the issue of Gaussian Bose sampling, “Jiuzhang” is millions of times faster than summit’s “Fuyue”, which is equivalent to 10 billion times faster than Google’s superconducting qubit computer “Suzuki”. < / P > < p > many people doubt that “Suzuki” solves the problem of random quantum circuit sampling, and “nine chapters” solves the problem of Gaussian Bose sampling, which is not a problem at all. How to compare? < / P > < p > Pan Jianwei expressed his understanding of such doubts: “it’s different things, how to compare them, like running and climbing. Therefore, in our article, we just say how many times faster it is. For example, Google’s task and our task are both 10 second algorithms. The fastest supercomputer takes 10000 years to calculate its task and 100 million years to calculate our task, which can be compared equivalently. This comparison is not particularly strict. ” Therefore, he called on the media not to leave the scientific connotation, otherwise it will not only have no role in promoting, but will bring negative effects. < / P > < p > Lu Pinyan, a professor at the school of information, Shanghai University of Finance and economics, also believes that the speed here is many times faster than that of the previous generation, which can not be understood by how much faster the next generation of computer chips are than that of the previous generation. < / P > < p > Lu Chaoyang added that the “nine chapters” quantum superiority experiment also overcame a loophole in Google’s experiment. “For example, its sprint is faster than the classic computer, but its long-distance running is actually slower than the classic computer. In terms of the existing theory, Jiuzhang can be faster than the classic computer, regardless of long-distance or short-distance running. ” < / P > < p > “nine chapters” does not mention the concept of “programmability”, which is remarkable. In response, pan Jianwei said that the quantum computer is programmable in principle, and team computing will be programmable in the next version of “Chapter 9”. < / P > < p > he explained in detail that before 2001, many people thought that the light quantum could only interfere in the interferometer and could not realize the strong coupling between light and light, so it was very difficult to make a logic gate. < / P > < p > “it’s very interesting that in 2001, the three scientists wanted to prove that it is impossible to do universal and effective quantum computation by means of light. As a result, in the process of proving, they proved that small quantum optical interference means are enough to do effective and programmable quantum computation. This caused quite a stir in our field at that time. ” < p > < p > Pan Jianwei said that the first quantum milestone was just to realize the superiority of quantum computing. “In the first milestone, we don’t care much about whether it can be programmed or not. We mainly care about whether it can be calculated quickly. For that task, whether it can be programmed or not is not very important.” < / P > < p > then, can the “nine chapters” be programmed in the future? “In fact, the modulation of light intensity, phase, refractive index and reflectivity can be programmable. We hope to do this in 2-4 years,” he said < / P > < p > in fact, both random route sampling of “Platanus acerifolia” and Gaussian Bose sampling of “nine chapters” are problems tailored for quantum computing, which can be said to be a “performance contest”. < / P > < p > in fact, shortly after Google announced the realization of quantum hegemony in 2019, IBM proved that through some optimization, the time for Supercomputing to solve the same random route sampling problem has been reduced from 10000 years claimed by Google to 2.5 days. < p >, can it be faster? Can the advantage of “nine chapters” resist the test of time? Even computer scientists have different views on this point. < / P > < p > Wang Feiyue, a researcher at the Institute of automation, Chinese Academy of Sciences, believes that the advantages of quantum computers are very reliable. At present, no computer scientist can prove that classical computing can solve such problems. He is more worried about the crisis caused by the rapid development of quantum computing. < / P > < p > and Lu Pinyan has confidence in classic computers. He said that Gaussian Bose sampling and random route sampling are approximate sampling problems corresponding to classical computation, but classical computers also have many good ways to complete computation without exponential time growth. Whether supercomputing can be faster is not a final conclusion. < p > < p > Pan Jianwei and Lu Chaoyang revealed that the next version of the “nine chapters” will expand 76 photons to more than 85 photons, thus increasing the original advantage of 100 trillion times by about 100 million times and widening the distance. < / P > < p > “quantum superiority is not achieved overnight, it is the continuous competition between quantum machines and classical computers. But I believe that quantum parallelism will eventually lead to capabilities beyond the reach of classical computers. ” < / P > < p > Pan Jianwei said that according to reliable statistics, Google has spent at least $1 billion on superconducting quantum computing in the past 10 years. Even within the team, the superconducting quantum computing research that Zhu Xiaobo is responsible for is “valuable”, and the whole process has to be processed by himself. “There are so many instruments made by the Academy of Sciences that we can definitely call.” < / P > < p > Pan Jianwei said frankly: “to be sure, in the short term, he has only spent money but not made money. Now is the stage of turning money into paper, and the era of turning paper into money has not yet come completely. ” < / P > < p > however, he assumes that if one day he suddenly discovers that Gaussian Bose sampling is a very important problem, and it can indeed calculate 80 photons, from this point of view, quantum computing will earn back all the money previously spent. < / P > < p > you know, in order to make a comparative verification on supercomputing, the research team of “nine chapters” helped out 400000 yuan of electricity bills and calculated 40 photons. If you really want to count 76 photons, the electricity bill will exceed the total output value of the United States. < / P > < p > of course, we have not found out the application value of Gaussian Bose sampling, which is difficult for Supercomputing and easy for “nine chapters”. < / P > < p > theory shows that a 20 million bit quantum computer can implement shor algorithm, a quantum algorithm of large number prime factorization, which may make the existing ciphers useless. < / P > < p > the more rigorous definition of “nine chapters” is actually the early quantum simulator. Pan Jianwei said that from his own point of view, he did not mind that the “nine chapters” were called quantum calculators or quantum abacuses. < / P > < p > Lu Chaoyang said that “quantum superiority” is the cornerstone of moving towards the ultimate goal healthily, and the second milestone after that is the special quantum simulator, which finds opportunities and provides some good solutions in the fields of material design, quantum chemistry, machine learning, etc. It takes hundreds to thousands of bits. < / P > < p > in recent years, Lu Chaoyang hopes that “Jiuzhang” can become a tool for experimental research, replacing part of the burden of supercomputing. Just like laser, it was also used for scientific experiments when it was first invented. < / P > < p > Zhang Shengyu, an outstanding scientist of Tencent quantum laboratory and tenured professor of the Chinese University of Hong Kong, does not think that non universality itself is any problem. If we can use a very special machine to solve an important problem, there is no problem in itself. < / P > < p > he said: “the key point is that this problem has good value in any aspect of scientific research or practical application. “Platanus acerifolia” and “Jiuzhang” have scientifically proved that the phenomena we observed on a relatively small number of particles before do not appear very different physics in multiparticle systems. It is also very difficult to achieve in engineering. Both teams are great, and I think they are undervalued in science and engineering. ” < / P > < p > although Zhu Xiaobo, Professor of the Department of quantum physics and quantum information, Shanghai Research Institute, University of science and technology of China, is not responsible for the research direction of optical quantum computing, he also thinks that people pay too little attention to the progress of science and technology in the “nine chapters”. He vividly said that the biggest significance of quantum superiority “is that children finally start to walk”, which is far beyond the meaning of who is faster and slower than others. Huawei has finally made a choice! Xiaomi and ov have also followed up. Have you ever thought about today?