Mr. Yao, the dean of our insitute, and Professor Yu made an expert interpretation of "Opinions of the State Council, the Central Committee of the Communist Party of China on Building a Data-based System to Give Full Play to the Role of Data Elements" ("Twenty Data Articles"), which was published in Guangming Daily on December 23rd.
The digital economy has entered a new era driven by data resources, and it is an inevitable requirement for economic and social innovation and development to cultivate the market of data elements and promote the circulation of data transactions. Based on the recently published Opinions of the Central Committee of the Communist Party of China and the State Council on Building a Data-based System to Give Full Play to the Functions of Data Elements, this paper considers and analyzes the core supporting technologies to promote the implementation of the data-based system based on the three core elements of digital economy: data, model algorithm and computing power.
The circulation transaction of data elements is the power engine of value highlighting. As a key element of digital economy, the circulation and transaction process of data is also a process of coordinated development and continuous progress of data, model algorithm and computing power, and its core is "taking data as resources and exerting information value". The data environment is a widely open environment, and it is in a state of frequent flow and sharing, which intensifies the cross-departmental and cross-industry data flow trend. Data has the characteristics of wide sources, large scale, rich structural types, diverse processing behaviors, separation of ownership and use rights, etc., which makes data face the risk of disclosure in an open environment.
Adhere to both data development and utilization and effective protection. On the one hand, it is necessary to legally circulate data and give full play to its potential; on the other hand, it is necessary to take necessary security measures to effectively protect data and ensure the ability to ensure a continuous security state. In recent years, the state has successively promulgated laws and regulations such as network security law, password law, data security law, personal information protection law, etc., to build a basic "compliance fortress" for network and data security, and to ensure data circulation transactions and digital development of the industry with data security, which will play a multiplier effect of data elements and promote the high-quality development of the digital economy.
The core elements of digital economy include data, model algorithm, computing power and so on. Data is becoming the key factor of economic production and the foundation of digitalization, networking and intelligence. It has been rapidly integrated into production, distribution, circulation, consumption and social service management, and has profoundly changed the mode of production, lifestyle and social governance. We need to study and promote the management of data confirmation and classification, smooth the flow of data transactions, improve the reasonable pricing of data, realize the market-oriented allocation of data elements, and rationally distribute the income of data elements.
Accelerate the cultivation of data factor market by means of innovative technology. Under the premise of complying with the requirements of personal information protection law, data security law and other relevant laws and regulations, we use secure multi-party computing technology to break the "data island" and realize the orderly circulation of data elements on the basis of data security and privacy protection. Explore the construction of data quality standardization system, accelerate the standardization of data collection and circulation transaction interfaces, and promote data integration and interoperability. Around the digital transformation, create a "data center" and fully integrate it under the premise of data security and data privacy compliance; At the same time, build cooperation alliances among enterprises, build an industry-level privacy computing platform, and form a data element circulation market among enterprises.
Secure Multi-party Computing (MPC) is the key technology of data security circulation. One of the biggest challenges of digital transformation is the problem of "data island": for different subjects, based on the laws and regulations of data security and data privacy, data cannot be simply integrated; For the same subject, data can be integrated to a limited extent through administrative orders, but it cannot be fully integrated by simple management requirements because of obstacles such as departmental interests, authority management and safety management.
Zero-knowledge proof (ZKP) is an effective supporting technology for market supervision of data elements. The data element market needs supervision and governance, which requires the construction of "supervision tools" and "supervision algorithms" on the premise of data security to ensure that the business activities and production activities on the data element circulation platform are carried out under the supervision framework and can be supervised and governed. The zero-knowledge proof cryptography can effectively achieve this goal. Government departments or industry alliances can take advantage of zero-knowledge proof technology to achieve both supervision and privacy, realize the public verifiability of data elements' compliance under the premise of ensuring data privacy, protect personal data privacy and business secrets of enterprises, and realize the goal of "the party manages data and the central enterprises take the lead".
Both supervision and privacy are realized. We will promote the data element circulation market among central enterprises to try first, and use zero-knowledge proof technology to achieve both supervision and data privacy. Referring to the successful experience of "supervision sandbox" in the financial industry, in the process of establishing the data factor market, the high-performance blockchain technology is used to realize the data's uplink storage and non-tampering, and at the same time, combined with zero-knowledge proof and other cryptographic technologies, the data privacy and data compliance can be audited and supervised.
Technology helps realize the market-oriented configuration of data elements. On the basis of data flow technical facilities established by secure multi-party computing, zero-knowledge proof, blockchain and other technologies, we will build a circulation market for data elements. These technologies can help realize data confirmation and classification management, and smooth the flow of data transactions. After solving the data circulation problems such as data security, privacy protection and data supervision, the most important thing is to form a reasonable economic distribution mechanism, so as to drive independent accounting within enterprises and data sharing among enterprises, so as to form a data element market, realize the market-oriented allocation of data elements, and rationally distribute the benefits of data elements.
In the era of digital economy, thousands of production activities in Qian Qian have spawned massive intelligent models. In recent years, artificial intelligence technology has made great progress, and the combination of big data and artificial intelligence makes it possible to build massive models. No matter in-depth algorithm research or horizontal innovation scene level, artificial intelligence technology has opened a magnificent curtain for the digital economy. Artificial intelligence model has strong advantages, which can be modeled without preset logic and expert opinions, and various algorithm tool platforms can support rapid modeling. However, the artificial intelligence model also has some shortcomings, mainly focusing on the interpretability of the model, the reliability of the verification method, the defects and limitations of the model and so on.
Algorithm model governance is the next chapter of artificial intelligence. The governance of the algorithm model, that is, the interpretability of the model, the accuracy and risk of the model, the fairness of the algorithm and so on. Model governance is both technology and system, and mature digital enterprises will form a complete model governance system. While promoting the construction of artificial intelligence capabilities, we should also promote the construction of mature model governance systems. At present, the research work including anti-attack, causal analysis and interpretable analysis has pointed out the development direction of this work. In these aspects, we believe that the following work should be done: First, establish a model governance system to form a mature system of model value evaluation, precision evaluation, risk and defect management. The second is to establish a model evaluation center, take the lead in building data model standards in combination with various scenarios, and practice model evaluation to empower the industry to develop.
The other side of model governance is the compliance of the algorithm. In the era of digital economy, more and more production activities need to rely on algorithm model, and the deep integration of algorithm and business will be the future trend, and the compliance requirements for business will also be transmitted to the compliance requirements for algorithm. At the same time, industry supervision is gradually being established. It is suggested that all industries should also establish corresponding industry norms and standards, and develop evaluation centers within the framework of laws and regulations such as data security law to test and evaluate the compliance and fairness of algorithm use.
Building a digital economy is inseparable from the support of computing power. The necessity of computing power in the process of digital economy construction is mainly reflected in five aspects: first, the data model needs to be realized on the computing power platform; Second, in the digital era of massive data, massive models and real-time computing, the demand for computing power has risen sharply; The third is to establish the circulation market of data elements, and the common modeling of multi-party data makes the requirement for computing power higher; Fourth, innovative software and hardware architecture, smart chips and future quantum computing technology all have great demand for computing power; Fifth, from the perspective of data element security, the algorithm is the main wrestling point between the offensive and defensive sides. Higher computing power support can effectively increase the cost of criminals stealing data, thus reducing security risks.
Layout a security system that "optimizes defensive efficiency". The essence of security confrontation is the war of attrition of resources between the enemy and ourselves, and optimizing the ratio of war losses between the enemy and ourselves is the key to victory. From the point of view of security technology, privacy computing technologies such as secure multi-party computing and zero-knowledge proof, while ensuring data privacy and security, have a certain degree of sacrifice in computing efficiency and communication complexity compared with similar computing without privacy protection. Therefore, in order to protect the security of data elements, it is necessary to comprehensively consider the advantages and disadvantages of various technologies, give priority to the integration and application of cryptographic technologies with "low computing power cost and high enemy consumption", adopt innovative data protection technologies without modification, and reasonably strengthen the investment in "protection computing power" to achieve effective data protection with lower cost and higher security.
Adhere to the independent control of key core technologies. Building a data-based system is the premise to support the digital economy, and the core technology is the support to promote and improve the establishment of the data-based system. Deep cultivation of technology, model construction and improvement of computing power are all indispensable components in the development of the digital economy. From a technical point of view, China should strengthen key core technologies such as secure multi-party computing, commercial cryptography and artificial intelligence, take the advantage of independent innovation, give play to the advantages of China's socialist system, new national system and super-large-scale market, improve the basic research and development capabilities of digital technology, win the tough battle of key core technologies, and firmly grasp the autonomy of developing digital economy.