On December 1, 2022, the Tencent Digital Ecology Conference “Techo Frontier Technology Forum” opened. Six scientists from Tencent and the Chinese Academy of Sciences discussed a number of cutting-edge technological trends at the forum, and shared the latest academic achievements and industry practices of Tencent AI Lab, Multimedia, Quantum, Tianyan, and Xuanwu Labs.
Among them, Zhang Shengyu, an outstanding scientist of Tencent and the head of Tencent Quantum Laboratory, issued an initiative to jointly build a “quantum +” application ecology at this forum, inviting partners in various fields to explore the application prospects of quantum computing in corresponding fields with Tencent, and promote While developing basic technology, it empowers the future of the industry.
(Picture: Zhang Shengyu, Distinguished Scientist of Tencent, Head of Tencent Quantum Lab)
At the forum, Tencent Quantum Lab demonstrated its layout ideas and related progress in the field of quantum computing. In the medium and long term, the Quantum Lab is committed to building a full-stack quantum computer technology, which serves as the basis for differentiated capabilities provided by Tencent’s external services. In the short term, the goal of the Quantum Laboratory is to research and explore the basic theory, realization path and application in related industries of quantum computing and quantum system simulation.
From 2021 to 2022, the Quantum Lab will publish nearly 50 articles, of which 15+ have an impact factor greater than 10. In terms of full-stack quantum computer technology, the laboratory has formed a certain accumulation. Recently, a multi-level high-parallel quantum control micro-architecture proposed by the laboratory and a scheme to suppress the ZZ crosstalk of quantum computers were respectively included in the computer top conference MICRO and ASPLOS; an efficient superconducting quantum computer proposed by the laboratory The bit initialization scheme and a superconducting qubit reading acceleration scheme were both published in the top international journal Nature Communications. At the same time, the quantum laboratory has carried out continuous innovation on quantum algorithms, covering long-term quantum algorithms that can be proved to be accelerated, quantum variational algorithms that are expected to be practical in the near future, and quantum adiabatic algorithms.
In terms of industry application exploration, Quantum Lab has actively combined self-developed cutting-edge algorithms and the massive computing power of Tencent Cloud to develop service platforms such as material research and new drug research and development, as well as open source software frameworks such as TensorCircuit quantum simulation, and has initially formed an application ecosystem. .
The following is the full text of the speech:
Hello everyone, I am Zhang Shengyu, the head of Tencent Quantum Lab. Today I would like to introduce to you some of the work of Tencent Quantum Lab.
Tencent Quantum Lab is a member of Tencent’s cutting-edge technology lab matrix. Our vision is to create world-class quantum computing science and industrial value, and our mission is to explore quantum technology and empower the future of the industry. Our work goal is to build a full-stack quantum computing system and related technological capabilities, and provide external services based on this to explore related industry applications.
At the lowest level, we provide powerful computing support with the help of Tencent Cloud’s massive servers. We realize the simulation of quantum systems and quantum algorithms based on classical computing power, and we are also building a full-stack quantum computer. On the second level, we conduct in-depth research on theories and algorithms including quantum algorithms, parametric subcircuits, AI+ science and AI pharmaceuticals, and strive to make world-class basic scientific and technological exploration work. The third layer, through the support of algorithms and computing power, we provide external platform services such as quantum computing, first-principle computing, molecular dynamics simulation, molecular database, virtual drug screening, AIDD tools, etc., for academic and industrial It can be used by relevant R&D departments in the world to help the common development of technology and industry. At the top level, we work with companies in materials, energy, semiconductors, medicine, finance, and information industries to study the specific applications of quantum computing in various fields, and strive to make quantum technology come true and generate real value.
Now back to the bottom. The full-stack quantum computer itself is a very complex system, and the full-stack quantum computer is an important capability base for us to provide computing services.
Next, we will talk about some progress in theoretical research and system development.
In terms of quantum computers, we proposed an efficient scheme for initializing superconducting qubits, and a scheme for accelerating the reading of superconducting qubits, both of which were published in the top international journal Nature Communications. In addition, we have also self-developed an impedance matching parametric amplifier, which has reached the domestic first-class level in terms of frequency, bandwidth and gain. In the middle digital-to-analog control and conversion layer, our self-developed measurement and control electronics system can support measurement and control of up to 90 qubits, and its volume and cost are much better than similar commercial instruments. We also proposed a high-parallel quantum computer architecture, with a performance improvement of 2 to 4 times, and the article was published on MICRO, the top meeting of the architecture. In addition, an encoder with a feedback loop and AI error correction is designed, which can be used for real-time error correction in the next step. On the PC side, we have proposed our own methods for noise characterization and suppression, which greatly improved the fidelity. The article was published on ASPLOS, the top conference. At the same time, it also innovates the compression method of quantum states and general quantum circuits, and improves the efficiency of quantum circuits.
In addition to the research on quantum computers themselves, we have also carried out continuous innovations in quantum algorithms. We have discovered quantum combinatorial algorithms with exponential acceleration, quantum AI algorithms with quadratic acceleration, quantum combinatorial algorithms based on dynamic programming, variable quantum algorithms that can more efficiently and accurately solve the ground state of quantum systems, and quantum adiabatic evolution accelerated by classical AI. Path design algorithm, etc. These articles were published in top conferences or top journals related to quantum computing, covering long-term quantum algorithms that can be proved to be accelerated, quantum variational algorithms that are expected to be practical in the near future, and quantum adiabatic algorithms.
We not only study quantum algorithms themselves from the perspective of mathematics and natural sciences, but also study corresponding quantum theories and algorithms based on real problems that arise in different industries. We have designed a quantum algorithm for molecular property prediction, an AI algorithm for molecular retrosynthesis prediction, a knowledge map prediction algorithm for drug-target interaction, a multi-objective optimization molecular generation algorithm, and an AI algorithm for molecular interaction adaptive graph learning, etc. . These results were published in the industry’s top journals or top conferences, and their performance significantly surpassed previous results, which enhanced our understanding of industry issues and the relevant scientific mechanisms behind them, and laid a solid foundation for future industry empowerment.
Next, I will introduce the work of Tencent Quantum Lab in application services and expansion.
New material research and biomedicine are important application areas of microphysics, in which first-principle calculations based on the quantum mechanics Schrödinger equation and simulation calculations based on molecular dynamics are important tools. Based on the powerful computing power of Tencent Cloud, we have developed a computing platform – TEFS, which supports multiple first-principles and molecular dynamics computing software. The platform has advantages such as elastic computing resources, friendly web interface, flexible Cloud Shell and efficient paper collaboration, and provides one-stop platform computing services for users in the materials and biomedical industries. Now we have formed cooperation with many leading universities and enterprises based on this platform to help the development of scientific research and industry.
In terms of new drug research and development, we have also launched a series of cloud services. Our virtual drug screening service can quickly screen candidate drugs that tightly bind to target proteins from a large library of small molecules. The druggability analysis service launched by us, combined with various self-developed algorithms in the laboratory, can provide users with the prediction of drug ADMET properties, thereby providing technical support for the druggability and safety evaluation of compounds in the early stage of drug development, and assisting the pipeline of drug development. Considering the problem of drug resistance of tumors and infectious diseases, we have launched a drug resistance database, which has greatly expanded the previous related databases in terms of sample size, protein type and mutation complexity. At the same time, we are also developing the structure prediction of mutant proteins, and look forward to helping advance and solve the problem of drug resistance from the aspects of data preparation and structure presentation.
Now back to quantum algorithms and software. Quantum algorithms can be roughly divided into two categories: those running on long-term large-scale error-correcting quantum computers and those running on medium-scale non-error-correcting quantum computers in the near future. The latter are mainly variational quantum algorithms. Many current demonstrations of quantum applications are actually concentrated in the latter category. Since the scale and quality of quantum computers still need to be improved, classical simulation has become an important method for the research of variable quantum algorithms at this stage.
We have launched the open source software TensorCircuit, a contraction engine based on the tensor network, which supports the three large machine learning libraries of Tensorflow, PyTorch and JAX. We can support heterogeneous hardware of CPU and GPU, and also support advanced engineering technologies such as automatic differentiation and just-in-time compilation. The software has a simple and easy-to-use API, which is suitable for scenarios such as variable quantum algorithms, quantum machine learning, quantum noise simulation, and simulation of ultra-large-scale quantum circuits, and its performance is far superior to similar products at home and abroad. For example, in terms of time, we run standard variable quantum algorithm tasks, which are orders of magnitude faster than Google’s TensorFlow Quantum and IBM’s Qiskit; in terms of space, for some typical one-dimensional quantum simulation tasks, we can simulate to The number of qubits is more than 600, which is more than 20 times the scale of the system that can be simulated by common state simulator software. We are committed to creating an open source ecosystem, and have also released relevant white papers, hoping to explore applications in different industries with our partners.
We work with partners in multiple specific industries to explore the use of quantum + high-performance computing to solve practical problems in the industry. In the field of biological computing, we cooperate with a leading company in this field on the development of protein reagents and MD computing tests. We provide partners with an optimized computing scheduling platform and GPU computing power, and assist partners in quickly testing tasks to complete product launches. In the field of drug development, we cooperated with a leading company to study quantum computing algorithms suitable for antibody H3 loop design. We used the algorithm idea of quantum walk to realize the parallelization of Monte Carlo sampling, proved that the proposed algorithm can speed up finding a stable antibody structure, and analyzed the specific time required. In the field of semiconductors, we cooperated with a head company of EDA self-developed system to jointly develop computing lithography software on the cloud and performance optimization, and jointly explored business models, completed the initial verification of SaaS products, and reached an agreement with end users preliminary cooperation intention.
The overall layout and related progress of Tencent Quantum Lab have been described above, including the computing power of classic high-performance computing and quantum computing at the bottom layer, quantum-related theories and algorithms at the middle layer, platforms and services provided, and top-level research in materials science, life science, Applications in the four domains of fintech and information technology.
Here we have issued an initiative to jointly build a “quantum +” application ecology, and sincerely invite partners in various fields to jointly explore the application prospects of quantum computing in corresponding fields with Tencent Quantum Lab, to promote basic technology research and development, and to empower the industry future.
This article is transferred from: https://www.leiphone.com/category/industrynews/8GshoAosyZA1BtX9.html
This site is only for collection, and the copyright belongs to the original author.