In-Depth | A new model of rule of law through generative artificial intelligence

What are your thoughts on the relationship between generative AI and the rule of law?

Professor Xu Juan:Generative AI and the rule of law are complementary to each other. Generative artificial intelligence has extraordinary language capabilities and has played an important role in many fields, such as medicine, education, and justice. It can be said that generative artificial intelligence is a starting point for the future development of the computer field and even the entire field. However, due to the complexity and application difficulty of generative artificial intelligence, there are many issues that need to be considered in the actual operation process. As far as the application of the rule of law is concerned, generative artificial intelligence can be applied in the production and promotion of legal knowledge, promoting the integration of the legal system and Science and technology develop simultaneously.

But this process also has two sides. Although generative artificial intelligence has advantages in handling many tasks, there are still challenges when facing complex problems. Therefore, we should pay more attention to how generative artificial intelligence aligns with the rule of law. When exerting its effectiveness, we should also pay attention to moral, ethical and other issues to ensure that emerging technologies can play a positive role in the rule of law. Technology should be used as the starting point for subsequent rule of law research. At the same time, laws and regulations must be formulated to ensure the development of technology and ensure that the law uses technology. To advance, technology has laws to follow.

What are your research ideas regarding the current application of generative artificial intelligence in the rule of law sector?

Professor Xu Juan:The core of my current research is the new model framework of the rule of law. Let me first introduce the specific meaning of three longitudes and nine latitudes.The three classics include the macro legislative knowledge background, which is divided into three dimensions: artificial intelligence foundation, application, and risk assessment; the microscopic legislative status facts are divided into three dimensions: data, algorithm, and computing power; the third classic is legislative norms, It includes three dimensions: purpose, principles and rules.

Generative artificial intelligence is multi-modal, imitating the thinking of the human brain and improving generalization capabilities. An existing application scenario of legal AI is mainly to provide enterprises with some intelligent legal affairs, intelligent legal question and answer, intelligent legal assistants and intelligent lawyers in some ordinary small claims litigation and legal text classification. Computer-interactive simulation scenes are presented. IBM has also developed an intelligent question and answer system in multiple fields of medical, financial and legal. There are also some small robots that can predict judicial trials in certain legal prediction fields and some risk prediction and warning systems for financial statement compliance in the financial field. These are applications at the micro level. Regarding risk assessment, first we need to set risk rating categories. In fact, there are already many risk assessment standards and typing standards, but risk rating categories and risk occurrence possibilities are different. Risk assessment is a composite process.

As a type of artificial intelligence, generative artificial intelligence has the same three elements as artificial intelligence: data, algorithms and computing power.Data is the foundation. A large amount of high-quality data does not have a huge amount of data. No matter how powerful the data computing system is, it is difficult for a clever woman to make a meal without straw. The three laws and three regulations on data also stipulate the quality standards of data, including identifiableness, security, sensitivity, etc. Algorithms are the way data generates value and is reconstructed. The most important thing about algorithms is transparency. Transparency is not only a computer issue, but also a political and legal issue. What are the consequences of opaque algorithms? Let me give you a very common example, that is, killing familiar users. The algorithm identifies old users and adopts different display methods for prices. I believe you have heard of this phenomenon more or less. This is the unfairness caused by the opacity of the algorithm. and human rights challenges. Computing power more reflects the company's technical hard power and is a driving force for data to be converted into value through algorithms.

Our legislation should address current difficulties. The law must not only reflect the value goal of security, but also respect the rights and interests of individuals, allowing everyone to independently choose which platform to use. This needs to be regulated by law and the state, and is also one of the goals of the law. The law initially evolved from ethics. The rule of law is actually a tolerant governance technique. The law does not tolerate the potential existence of bad things, but does not tolerate the explicit stimulation of bad things. It is a bottom-line standard.Generative AI cannot actually be counted as "generative" because it does not actually create new knowledge and content. It should be defined as "auxiliary" AI. Our laws should regulate the application of this technology. Of course, this does not refer to the low-end technical issues of the technology itself.

What changes do you think the new rule of law model of generative artificial intelligence will make to the existing rule of law model?

Professor Xu Juan：There are currently three main approaches to the construction of artificial intelligence legislation, including legal availability + improved legislation, social response + exploratory legislation, and technology transformation + integrated legislation.The first is improved legislation. In fact, the phenomenon of technology improving the rule of law has always existed. For example, without the emergence of the Internet, the Personal Information Protection Law would not have been produced. The emergence of new technologies will inevitably bring new challenges and opportunities to the law. I have also listed 12 laws that can be improved through new technologies, including the legal rights system, the legal obligation system, the property rights legal system, the legal governance system, and the legal system. Procedural system, legal liability system, legal discretion system, legal supervision system, risk legal system, evidence legal system, legal power system, and political and legal traditional system. Of course, it is impossible for technology to completely subvert the law, and it is impossible to establish a new legal system to replace the existing legal system. It is more about some work that can be done at the level of improvement and perfection. Think about the direction of improving existing laws from three perspectives: data, algorithms, and computing power.

Second exploratory legislation. Law should be socially interactive, responsive to social scenarios, and a technological transformation experiment to improve social responses. For example, on the issue of people with disabilities, urban planning needs to provide them with some meaningful digital services and demonstrate an attitude of respecting differences and diverse perspectives. We should use the rule of law to encourage the expansion of human knowledge and capabilities, use some functional designs to consider the key needs of different groups of people, use technology to promote and regulations to guide, so that functions can be implemented.

The third integration legislation. In fact, the main problem here is to deal with the problem of wrong information, such as inaccurate input and output, false information, etc. Based on this problem, it can be divided into two steps. The first is to distinguish between trainable general legal thinking and untrainable decision-making legal thinking. For example, some legal language can be expressed and interpreted by machines, but there are some terms that are difficult for machines to recognize. The second is to transform languages that are difficult for machines to recognize. This kind of structural language ability needs to be improved experimentally. However, it should be noted that even after language training, it is difficult to ensure that the machine can make a responsible judgment. The understanding of law is not a mechanical process, it is a dynamic input into human vitality that is injected into every specific issue in a specific scenario.

Regarding the risks of generative artificial intelligence products - such as creating false information, infringing on the rights of others, etc., have you noticed any specific measures?

Professor Xu Juan:I think the more clichéd answer to this question is that technology needs to be optimized. Improve the accuracy and credibility of algorithms, and the transparency and revealability of generalization ability models. There are also risk prevention and control measures and content review steps for products. There are also joint supervision of ethics committees and improvement of quality and literacy, etc. There are many measures in terms of measures.

The application of generative AI may have an impact on the job market and labor demand. How to deal with possible social and economic problems through legal means?

Professor Xu Juan:This question reminds me of an interesting phenomenon. Recently, large model clauses have become popular. I have seen some legal scholars post a lot of relevant opinions, which involve large platform individuals as stipulated in Article 58 of the Personal Information Protection Law. A summary of the norms governing information processing obligations, they are known as gatekeeper clauses. So what does this clause have to do with market employment? In fact, it can be understood this way. For example, the developer of a large model is usually the gatekeeper of digital technology, and he has great market power. Then this dominant power can actually produce market benefits. A good platform fulfills many obligations. Relatively speaking, the gatekeepers of the platform have certain means to deal with this social and economic problem. We have also talked about generative artificial intelligence before. It is only an auxiliary tool and cannot replace the human brain for thinking. Coupled with the existence of gatekeepers on large platforms, it is impossible for people to be kicked out of the job market. I have always emphasized the noble souls of people. The only way to solve employment pressure from the root is to make continuous progress and improve quality. People must catch up with the pace of social development, because the progress of society is objective and necessary, and the law does not May do perverse things. So this is actually not a technical issue.

Is there anything else you would like to say to everyone?

Professor Xu Juan:Today we are in an era of anti-globalization, anti-globalization and narrow nationalist extreme thoughts, provoking a Cold War mentality and confrontation between Eastern and Western civilizations. The international situation is pursued in some countries and regions. Under the confusing circumstances, global consensus is impossible. As scholars, we should keep a clear mind. For example, Yin Haiguang once said that intellectuals are the eyes of an era. The eyes of intellectuals should not and cannot be obscured by the various turbid currents of this era. Historically, the thinking and construction of international order often occurred when the international situation was dark and unclear.

International organizations and forums develop recognized standards that apply to all participants. The standards reflect that the international community is slowly dividing AI ethics into different levels, types, and standards. Looking for consensus in different industries. Of course, consensus is difficult to reach in some industries. It is a moral responsibility for scientists to intervene and multinational companies to take the lead in setting an example and jointly reasonably guide and effectively support the development of a long-term project that is conducive to improving the well-being of all mankind. To guide and promote countries to realize the global construction plan of AI ethics as soon as possible. One of the biggest problems we face today is narrow-mindedness. Civilization under the rule of law is a high-level civilization, and the new model of the rule of law needs to eliminate narrow-mindedness, unfairness, hegemony and discrimination. The problem that generative artificial intelligence needs to solve is to expand language clarity without creating new language confusion, and to allow large and small language models to coexist harmoniously. The national differences and global consensus on robot legal culture are actually our appeal to human consciousness when technology and business develop together in a smooth society like ours today. It is a booster of the new language of safeguarding the constitution and using administration to protect, lead and expand technology across domains from two dimensions.