Artificial intelligence legislation is about to come out, how will it drive technological innovation and optimize the industrial structure

Cai Shuyue, reporter of the 21st Century Business Herald, reported from Shanghai

In recent years, with the continuous iteration of AI technology and the rapid development of generative AI applications such as ChatGPT, to a certain extent, countries around the world are also forced to introduce supporting policies and bills as soon as possible to standardize the path of technology research and development, and try to prevent ethical and safety risks and compliance risks that may occur in artificial intelligence products and services, and promote the positive and stable development of the entire industry chain.

At present, the legislative work of artificial intelligence in the mainland is gradually advancing. Recently, the General Office of the State Council issued the "State Council 2023 Annual Legislative Work Plan", which mentioned that in the implementation of the strategy of rejuvenating the country through science and education and promoting cultural self-confidence and self-improvement, it is preparing to submit the draft artificial intelligence law.

After combing the reporter of the 21st Century Business Herald, it was found that the current mainland artificial intelligence legislation adopts a more diversified regulatory approach, such as formulating special artificial intelligence regulations, adding artificial intelligence-related provisions to relevant laws, and adding artificial intelligence-related content to the management regulations in related fields.

In addition, local policies are also a major feature of the mainland's artificial intelligence legislative process. For example, Beijing, Shanghai, Shenzhen and other places have issued local policies or guidance documents related to artificial intelligence.

Some interviewed experts pointed out to 21 reporters that from the current regulatory attitude of the country, it can be seen that there is a conscious "tailor-made" design for a certain industrial field and a specific technology extension block. Detailed to the localities, local governments consciously raise the priority of industrial independent development, formulate personalized development plans according to the specific development status of the industry, and take into account the risks of industrial development, which are also specific actions for the mainland to support the development of the artificial intelligence industry.

The pace of legislative exploration has accelerated

The beginning of the mainland's AI-related legislative work can be traced back to the "New Generation Artificial Intelligence Development Plan" issued by the State Council in 2017. At that time, the strategic goal planned in this document mentioned that in 2025, artificial intelligence laws and regulations, ethical norms and policy systems should be initially established to form artificial intelligence safety assessment and control capabilities.

Peng Kai, a senior partner and lawyer at Jincheng Tongda & Neal Law Firm, pointed out to 21 reporters that the current artificial intelligence legislation in the mainland is generally in a framework state of "exploration + acceleration".

In the "exploration" link, Peng Kai believes that the mainland's current integrated artificial intelligence-related legislation is not prominent, and the legislation with a higher legislative level and the nature of domination has not yet been announced, and it is more scattered in multiple fields and throughout the multi-industry chain, and it is not difficult to see the intention of exploration.

In addition, compared with the situation of countries around the world, there are fewer countries with reference legislation in the field of artificial intelligence, and all countries in the world are crossing the river by feeling the stones, and many countries do not have a regulatory document specifically for artificial intelligence.

The European Union, one of the current pioneers in the development of AI-related rules, first made public the AI Act in 2021 and approved by the European Parliament on June 15. It is reported that the draft artificial intelligence law passed by the vote will ensure that artificial intelligence developed and used in Europe is fully in line with the rights and values of the European Union, including human supervision, safety, privacy, transparency, non-discrimination, and social and environmental well-being.

At present, there is no systematic artificial intelligence legislation in the United States, and the relevant strategies mainly rely on executive orders and guiding documents to promote. For example, on May 23, the White House released the National Artificial Intelligence R&D Strategic Plan, which in addition to updating the 2016 and 2019 versions of the National Artificial Intelligence R&D Strategic Plan, also proposed to evaluate federal agencies' response to the National Artificial Intelligence Initiative Act of 2022 (NAIIA) and the National Artificial Intelligence R&D Strategic Plan The specific implementation of the situation.

Looking back at China, with the rapid development of artificial intelligence technology from 2022 to 2023, the mainland's artificial intelligence-related management policies have sprung up like "mushrooms" in the past two years, and management policies in many subdivisions have been introduced successively. For example, the Provisions on the Administration of Internet Information Service Algorithm Recommendations (January 2022), the Provisions on the Administration of Deep Synthesis of Internet Information Services (November 2022), and the Measures for the Administration of Generative Artificial Intelligence Services (Draft for Comments) (April 2023) respectively put forward corresponding requirements for technical compliance from different levels such as algorithm governance, deep synthesis governance, and generative AI governance.

Peng Kai also mentioned that in terms of the current effective type of artificial intelligence legislation in the mainland, most of them are draft for comments or are still in a trial state, and the tentative significance is high; From the perspective of the type of legislation, there are more documents with a forward-looking narrative style, such as guiding principles, planning plans, and action plans, which are more pronounced and advocated; From the perspective of the type of application of legislation, the current legislation is more inclined to promote development with high fault tolerance, while there are fewer punitive and regulatory legislation similar to the trial and error cost.

Full-cycle governance of industrial development

Since the beginning of this year, major domestic Internet companies have opened the AI big model "horse racing", and the large models that have been announced so far include Baidu Wenxin big model, Ali Tongyi big model, Huawei Pangu big model, etc., all of which are based on deep learning algorithms, and have made significant progress in natural language processing, image recognition, speech recognition and other fields.

On the one hand, the rapid development of the AI industry has directly accelerated the exploration process of relevant legislation in the mainland, and also changed the previous regulatory logic.

Zhang Xin, executive director of the Digital Economy and Law Innovation Research Center of the University of International Business and Economics, pointed out to 21 reporters that two regulatory documents, such as the "Provisions on the Administration of Internet Information Service Algorithm Recommendation" and the "Provisions on the Management of Deep Synthesis of Internet Information Services", are more like the rules set by the "1.0 era" of artificial intelligence, "and the "Measures for the Management of Generative Artificial Intelligence Services (Draft for Comments)" are the development of artificial intelligence in the mainland." The 2.0 era' was announced, and the industrial chain status and regulatory logic it faced have undergone great changes. She said.

The layered nested industrial chain increases the complexity of supervision. If the AI field is segmented from the industry chain level, it can be divided into the basic layer, the technology layer and the application layer.

Among them, the basic layer mainly includes the underlying training data and data centers, servers, chips, etc. that provide computing power, the technical layer includes natural language processing, deep learning, machine learning and other algorithms and large models equipped with the above algorithms, and the application layer is the specific scenarios of artificial intelligence, such as ChatGPT, Midjourney, Wen Xin Yiyan, Tongyi Qianwen, etc. that have emerged in recent months.

"In the context of the gradual clarification of the technical framework of artificial intelligence, from the basic layer to the application layer, it is necessary to examine the corresponding risks from a full-cycle perspective." Zhang Xin pointed out that AI is essentially a "complex", and its legislation is closely related to the algorithms, data and platform governance behind the technology.

Therefore, how to achieve the combination of agile governance and resilient governance in the legislative process is a challenge for both regulators and industry.

On the other hand, Peng Kai pointed out that from the current regulatory attitude of the country, it can be seen that it is consciously "tailored" legislative design for a certain field and specific technology extension. Specific to the industrial chain where various artificial intelligence may be applied, such as medical treatment, autonomous driving and other fields, Continental has issued corresponding specifications such as the "Management Code for the Clinical Application of Artificial Intelligence-assisted Treatment Technology (2022 Edition)" and "Classification of Automobile Driving Automation". "Therefore, our current legislation considers risk points relatively comprehensively, and even if it cannot be accurately anchored, we try to incorporate them into management." He said.

Local policies promote independent innovation

In addition to the layout of enterprises, many provinces and cities have also launched an artificial intelligence "arms race". Recently, the "Chinese Intelligent Big Model Map Research Report" released by the New Generation Artificial Intelligence Development Research Center of the Ministry of Science and Technology shows that taking the regional distribution of domestic large models as an example, Beijing, Guangdong, Zhejiang and Shanghai are in the first echelon, of which Beijing and Guangdong have 38 and 20 large models under research respectively.

In the above context, local governments are also simultaneously promoting the implementation of artificial intelligence policies. For example, Beijing, Shanghai, Shenzhen, Hubei and other places have successively introduced relevant plans, all of which show that local attention and attention to the artificial intelligence industry.

Peng Kai observed that these local AI-related policies have one thing in common, that is, they highly emphasize the development of industrial autonomy. He believes that in the future, as an industrial blue ocean, it is an important foothold for the field of artificial intelligence to seize development opportunities, break through key technologies, and enhance their core competitiveness. This is not only a requirement for competition between cities and regions, but also the basic premise for the country's artificial intelligence development process to be at the forefront of the world.

For example, Beijing pointed out in the "Implementation Plan for Accelerating the Construction of a Source of Artificial Intelligence Innovation with Global Influence (2023-2025)" issued by Beijing on May 30 that between 2023 and 2025, the market share of basic software and hardware products such as domestic artificial intelligence chips and deep learning frameworks will be significantly increased, and computing power chips will basically achieve independent and controllable goals.

Shenzhen, on the other hand, mentioned in Article 15 of the Regulations on the Promotion of Artificial Intelligence Industry in the Shenzhen Special Economic Zone that it will carry out strategic, forward-looking and systematic basic research on artificial intelligence and key core technology research, promote breakthroughs and innovations in discipline theories and cutting-edge technologies, and play a leading and supporting role in innovation.

"Breaking through the 'stuck neck', abandoning 'relying on imitation', and adhering to 'independent innovation' are the development spirit that our provinces, cities and countries should uphold in the field of artificial intelligence in the future." Peng Kai pointed out.

On the other hand, the "personalized" characteristics of local regulations also reflect their focus on the development characteristics and objective development of local industries. "For example, Beijing and Shanghai are positioned as international science and technology innovation centers, while Shenzhen is defined as a new type of special economic zone and economic development innovation place. When specifying relevant policies, local governments will also carry out point-to-point design tailored to local industries. Zhang Xin pointed out.

However, the AI threat theory has always been the focus of high attention in the legislative field. Tesla CEO Elon Musk has bluntly said that "the risk of artificial intelligence is much higher than nuclear weapons", and in late March, an open letter signed by more than a thousand entrepreneurs and scholars "Pausing large-scale artificial intelligence experiments" also pointed out that no one, including the developers of the technology, can really understand, predict or fully control the technology.

Therefore, while encouraging technological innovation, local governments are also taking into account the potential risks and ethical issues. For example, Beijing mentioned "improving the ethical governance capacity of artificial intelligence science and technology." Strengthen research on AI ethical safety norms and social governance practices", Shenzhen also pointed out that the development of artificial intelligence should be "safe and controllable" and "should comply with artificial intelligence ethics and safety norms".

"In the final analysis, because of the emergence of artificial intelligence, it is equivalent to adding an additional 'new person' between the original 'people' - a 'person' that the previous law failed to cover. If one day in the future, artificial intelligence truly achieves 'thinking and acting like a human', will we still be able to use it as a tool, a machine? This is a huge challenge not only for the continent, but also for countries around the world. Peng Kai shared his thoughts.

For more information, please download 21 Finance APP