Bosh Allen report: China's quantum computing will surpass the United States and Europe
Recently, Booz Allen Hamilton, an American consulting firm, released a 32 page report, which mentioned that quantum computing may pose a threat to American information security, especially worried about China's progress in the field of quantum computing, and urged the chief information security officer (CISO) of American companies to prepare in advance.
Although the title of the report is "machine threads in the quantum era", most of the content summarizes the important progress in the field of Quantum Information Science in China. "Since 2010, the Chinese government has publicly identified quantum computing as a key strategic technology for its economy and national security, and has successfully narrowed the gap with the United States through talent, capital and policy support. Based on the current trend, China may surpass the United States and Europe." finally, the report also puts forward a roadmap for the application of quantum computing.
According to the report, China was not a major participant in the early research and development of quantum information science. Then, from about 2016, China quickly caught up and announced major technological breakthroughs and considerable government and commercial investment.
In 2016, the Chinese government announced the 13th five year plan, which lists scientific and technological innovation as a national core priority, aiming to make China a global innovation leader in 2030. To this end, the plan requires the creation of major scientific and technological projects in six fields, including quantum computing and communication research. These projects aim to build the national quantum communication infrastructure, develop a general quantum computer prototype and build a practical quantum simulator.
China's 13th five year plan has brought huge investment, especially for the establishment of special laboratories and quantum R & D teams.
In 2017, China announced plans to build a large-scale National Laboratory of quantum information science at a cost of US $11 billion. Also in 2017, the state-owned enterprise Anhui Investment Group established a $1.5 billion fund to invest in Chinese companies developing quantum technology. China's major technology enterprises have followed suit. In 2017, Alibaba announced that it would invest US $15 billion in "basic and disruptive technology research" in the next three years, including data intelligence, natural language processing, machine learning and quantum computing, and established seven laboratories in China, Russia, Israel and the United States.
Although China has made great progress in quantum information science, many Chinese scientists believe that China lags behind the rest of the world. In October 2020, Xue Qikun, then vice president of Tsinghua University, made a report on China's quantum development at the 24th collective study on the research and application prospects of quantum science and technology held by the Political Bureau of the CPC Central Committee. He pointed out that although significant progress has been made, China's quantum development faces many deficiencies and challenges in terms of policy and investment support, multidisciplinary integration and training. In the same month, Guo Guoping, China's top quantum scholar, told the global times that China is in a leading position in the field of quantum communication, but it is 3-5 years behind in the field of quantum computing. He believes that China's scientific research and its development in quantum simulation and other practical applications are insufficient.
The report summarizes China's important progress in four quantum fields in recent years, including secure communication, computing power, quantum stack and quantum assisted artificial intelligence.
1. Secure communication
Quantum encrypted communication is basically the same as classical encrypted communication. Keys are exchanged between the two sides, so that both sides can encrypt and decrypt mutually understandable data. Quantum key distribution (QKD) is different from key exchange on classical computers because keys are established using quantum channels. Due to the role of quantum properties, any attempt to intercept the key will arouse the vigilance of both sides of the communication and refuse to decrypt the eavesdropper without authorization. Although the U.S. Department of Defense (DoD) conducted some early tests on QKD, the United States rejected this technology to a large extent. For various security and cost reasons, the U.S. National Security Agency (NSA) does not encourage investment in QKD. It is publicly evaluated that the development and implementation of anti quantum cryptography technology is a "more cost-effective and easy to maintain solution" than QKD.
The more practical use of quantum networks may be distributed quantum computing. For example, the EU has invested 1 billion euros in a project called "quantum flagship", which aims to enable quantum computers across the European continent to communicate with each other.
China's first major quantum information science milestone is related to communication security. Since then, China has become a leader in this field. China has explored some transmission mechanisms of QKD and established large-scale infrastructure in the densely populated east.
Satellite link
In 2017, China reported the first successful QKD through a satellite named after the ancient Chinese philosopher Mozi. The original system was mainly used for demonstration purposes. It is very error prone and requires that the key can only be sent at night to reduce the impact of sunlight. In the same year, China reported that it had made the first intercontinental quantum encrypted video call with the president of the Austrian Academy of Sciences in Vienna using the Mozi satellite.
In 2020, China launched the world's first mobile quantum ground station, which successfully communicated with Mozi for nearly 8 minutes. The equipment weighs only 80kg and can be installed in the vehicle. China's 14th five year plan (2021-2026) lists satellite quantum network as a national key research topic.
Ground link
In 2017, China launched a network with more than 2000 kilometers of dedicated optical fiber for quantum communication, connecting governments and state-owned commercial entities in Beijing and Shanghai, as well as intermediate cities Jinan and Hefei. The network adopts the relay principle. The quantum key must be converted into classical data at 32 intermediate nodes (repeaters), and then converted back to quantum data for retransmission.
Quantum communication "Beijing Shanghai trunk line"
In 2018, researchers at Tsinghua University successfully entangled 25 quantum interface links (also known as quantum repeaters), far exceeding the existing records. This repeater uses entanglement for retransmission without classical data conversion. In 2021, Chinese scientists reported that without repeaters, quantum keys were sent over 500 kilometers through optical fibers.
Hybrid network
As of January 2021, the Beijing Shanghai quantum secure communication network has expanded to about 4600 kilometers, mixed with optical fiber and ground to satellite links, and supports more than 150 users. The satellite link has established a connection between the east coast of China and Urumqi in Western China.
This network constitutes the largest stable QKD network in the world. Although the transmission rate is still very slow, about 5 Kbps, QKD only transmits small keys through quantum channels and sends a large amount of encrypted data through traditional methods. China's goal is to establish a global QKD network by 2030.
2. Computing power
In December 2020, China claimed the superiority of quantum computing for the first time. In an article in science, a research team from several Chinese institutions reported that they had built a computer called "Chapter 9", which can complete a specific calculation in a few minutes, while the world's most powerful supercomputer takes more than 2 billion years to complete.
In 2021, China announced the successful development of 113 photon "Jiuzhang 2" quantum computing prototype. According to the officially published optimal classical algorithm theory, "Jiuzhang 2" can process Gaussian Bose sampling 1024 times faster than the fastest supercomputer at present. Meanwhile, the 66 bit programmable superconducting quantum computing prototype "Zuchongzhi 2" realizes the "quantum computing superiority" of the superconducting system, and the computational complexity is 6 orders of magnitude higher than that of Google "sycamore".
Zu Chongzhi 2 quantum processor
3. Quantum stack
A new software stack must be developed to take advantage of the capabilities of quantum computers. Generally, these stacks are regarded as hybrid classical computers interacting with quantum processors, just as many classical computers today use GPU to perform special calculations for rendering graphics. There is no consensus on which specific components exist in the quantum stack.
In February 2021, a Chinese company called origin quantum released the first quantum operating system in China, called origin pilot. Native quantum indicates that their operating system is superior to the quantum operating system developed abroad (such as deltaflow.os in the UK and parityos in Austria). It is reported that Benyuan Sinan is the first quantum operating system that allows multiple processes to run in parallel (i.e. multi task parallel), and has uncommon functions such as automatic calibration, which can improve the stability of the system.
4. Quantum assisted artificial intelligence
The practical limitations of data processing on classical computers are obvious. In recent years, AI has excelled in gaining new insights from large data sets. Although larger data sets can produce more insights, they also require more time to process.
According to norishige norimoto, vice president of research and development of IBM, "the large amount of data generated in the 21st century needs to double the computing power every year, and the cost and energy consumption are too high to maintain." despite these challenges, the continuous economic and national security pressure requires more in-depth insights from New and growing data sets, which requires new data processing and analysis methods.
Quantum computing can solve this big data problem. For example, quantum computers can accelerate machine learning by creating weighted models for interpreting data.
In recent years, China has vigorously developed its artificial intelligence capability. In 2017, the Chinese government announced the national AI development strategy - a new generation of AI development plan. By 2020, the scale of AI core industries will exceed 150 billion yuan, and AI technology will reach the world leading level by 2030. In 2019, the Ministry of science and Technology issued the governance principles of a new generation of artificial intelligence - developing responsible artificial intelligence. Since 2015, the scale of China's artificial intelligence market has increased by 44% every year, from an estimated US $2 billion to US $11 billion.
Some progress has been made in the development of quantum machine learning in China. In 2018, Baidu, a large technology group, announced the establishment of Quantum Computing Institute, focusing on quantum artificial intelligence and information science. In May 2020, the company released the first open-source machine learning and deep learning development kit for quantum computers in China - pad quantum. It supports major quantum computing research topics such as machine learning, chemical simulation and combinatorial factor decomposition.
Finally, the report lists the main quantum information technology research teams in China. From north to south in the map are Baidu Institute of quantum computing, Alibaba Quantum Computing Laboratory (Beijing), Alibaba Quantum Computing Laboratory (Shanghai), ZTE (Shanghai), Shanghai Branch of Chinese Academy of Sciences, source quantum, GuoDun quantum, National quantum information science laboratory University of science and technology of China, Alibaba Quantum Computing Laboratory (Hangzhou), Tencent. For more information about China quantum technology company, you can see the photon box overview of domestic quantum technology companies (2021 update).
The following figure shows the time prediction of the report for quantum computing applications in three fields, from top to bottom: simulating complex systems with quantum characteristics, cracking public key encryption and bringing advantages to artificial intelligence. Among them, the grid rectangle indicates optimism, the slash rectangle indicates consensus, and the red rectangle indicates pessimism.
1. Near term prospects of nisq computing (2023-2040)
Nisq is a medium scale quantum with noise. This concept was proposed by Professor John Preskill of California Institute of technology. Nisq computer is very suitable for studying systems with quantum properties, also known as quantum simulation. Over the past decade, established and start-up pharmaceutical companies have shown great interest in the use of quantum computing in drug discovery.
Since there are almost no error correcting qubits on the nisq computer, researchers propose to use the "hardware efficient" combination of quantum and classical hardware to solve the problem, in which the quantum algorithm only focuses on the most difficult aspects of classical modeling. For example, in the case of chemical problems, the strategy will use quantum computers to prepare test states and measure relevant expectations. Among them, quantum computer solves the most difficult part of classical simulation calculation——optimization problem, rather than the whole simulation in quantum environment.
This computing strategy is similar to the CPU and GPU of a classic computer. All classic computers have a CPU, which is a general-purpose processor. Some classic computers also use GPU, a special processor used to deal with specific computing problems, such as fast drawing graphics, deep learning and mining cryptocurrency. Similarly, this computing strategy manages the problem that quantum computer errors increase with the increase of the number of operations by limiting the number of algorithms running on quantum computers.
When is the nisq computer used for quantum simulation? Running quantum simulation requires about 150 error correcting qubits, and each qubit requires dozens to thousands of physical qubits for error correction and stability. Although the nisq computer will soon become powerful enough in this regard, the trend of rapid collapse of qubits and high error rate greatly limits the complexity of evaluable problems. In addition, with the increase of the number of entangled qubits, the possibility of noise or collapse increases.
By the end of the 1920s, it will be possible for the nisq computer to run quantum simulations under various conditions.
● at present, the most optimistic estimates are Goldman Sachs and QC ware, a quantum computing as a service company. They predict that quantum computers will be used for chemical simulation from 2024 to 2026. Bosh Allen has reservations about this prediction.
● by the end of 1920s, quantum computers will generally perform better than classical computers in quantum simulation. For example, the EU's quantum flagship program estimates that by about 2030, prototype quantum simulators will surpass supercomputers in fields related to chemistry and material science.
● more pessimistic forecasters such as Boston Consulting Group (BCG) point out that more optimistic estimates are highly dependent on some companies and researchers to achieve the expected technical milestones, which means that quantum simulation may not be realized until 2040. They pointed out that the development of small molecule simulation on classical computers has matured, which means that quantum computers must be fully mature before showing performance improvement.
● the Chinese government is optimistic about the prospect of quantum computing in the next few years. In the 14th five year plan, China set the goal of 2025, that is, to have a quantum simulator that can surpass the classical computer in solving several different problems and manipulating hundreds of entangled qubits. China's top scientists point out that China lags behind the United States in quantum simulation.
2. Long term prospects of general quantum computing (2030-2050)
The general fault-tolerant quantum computer is superior to the classical computer in advanced simulation, search and optimization, but it will take a long time to develop. Although most nisq computers are generic, their inability to correct errors limits them to solving problems with few computational steps. Before waiting for other solutions, error correction will need to create abstract error correction "logical" qubits, each containing thousands of physical qubits. Achieving this will be a huge engineering challenge.
In general, the universal fault-tolerant development milestone is expected to be reached in the 1930s. Google and IBM aim to develop general-purpose quantum computers by 2029 and 2030, respectively. BCG estimates that this milestone will be achieved sometime between 2031 and 2042. In December 2020, Chinese quantum computing scientists also estimated that China would develop a general-purpose quantum computer between 2035 and 2040.
The National Academy of Sciences, the Academy of engineering and the school of medicine believe that this milestone is so far away that it is unpredictable. They believe that the difficulty of this technical challenge cannot be evaluated due to the lack of existing research on how to connect quantum modules to create large-scale quantum computers (> 1000 fault-tolerant qubits).
Once the large-scale fault-tolerant quantum computer is built, all public key cryptography algorithms widely used today will be vulnerable to attack. The report cited a paper earlier this year, saying that cracking rsa-2048 requires 4098 fault-tolerant (logic) qubits or 20 million noisy qubits. When can such a computer be built? The earliest estimate is usually around 2030, while the most common conservative estimate is 2040-2050.
It is worth mentioning that the latest research shows that 20 million physical qubits may not be required to crack rsa-2048. Researchers from the French new energy and Atomic Energy Commission have proved that 13436 qubits can crack rsa-2048 by integrating the quantum memory into the quantum computer, which is three orders of magnitude less than the number of qubits required in previous research.
3. Further development of quantum computing (time unknown)
During this period, the application envisaged in the report was quantum assisted artificial intelligence. Quantum assisted artificial intelligence requires a lot of engineering and computer science progress to become useful. Some researchers believe that although even ordinary nisq computers theoretically have enough computing power to perform machine learning, the amount of error correction required makes them unusable in practice. Therefore, artificial intelligence on quantum systems may require a large number of unknown highly stable or error correcting qubits. In addition, in order to convert a large classical data set into data stored on qubits, it is likely to be necessary to create a new form of memory tentatively called QRAM.
Finally, before quantum assisted AI is considered valuable, a new quantum assisted AI model must be developed, which is significantly faster than the classical AI model. As stated in a paper published in the journal Nature in January this year, "quantum computers may be able to speed up complex computing on small data sets, but they have less advantages in solving problems on large data sets."
Considering a large number of outstanding issues, it is unclear when quantum computing will have a meaningful impact on artificial intelligence.
In 2021, a group of researchers from several universities in India found that quantum computing has great prospects for promoting the development of artificial intelligence and machine learning, but the relevant research is mainly theoretical, and there are few experiments to prove how it will be applied in practice.
The EU's quantum flagship program estimates that quantum computers will outperform supercomputers in optimization problems related to artificial intelligence by about 2030.
Link：https://www.boozallen.com/expertise/analytics/quantum-computing/chinese-cyber-threats-in-the-quantum-era.html
