[author]Zheng Ge

[content]

Zheng Ge  Rethinking on the Future of Artificial Intelligence and Law

*Author Zheng Ge

Professor at Kaiyuan Law School of Shanghai Jiao Tong University

Director of the Planning Committee of the Chinese Academy of Law and Social Studies at Shanghai Jiao Tong University

Abstract: Artificial intelligence is often envisioned as a technology that replaces humans. In the legal field, robot judges and robot lawyers often raise concerns, especially with the breakthrough development of generative artificial intelligence causing a new round of anxiety about "machines replacing humans". But human-machine symbiosis and human-machine interaction are the common phenomena now and in the future. Starting from human-computer interaction and developing relevant rules around the principle of "people in the loop", we can effectively prevent and control risks and actual damages of artificial intelligence, and fully utilize technology to assist and strengthen human work, including legal work, so that artificial intelligence can serve the purpose of enhancing human well-being.

Artificial intelligence has become a universal technology in our era. The so-called universal technology refers to a technology whose use is not limited to a specific field, but can be applied to multiple fields (including areas that humans have not yet thought of). This technology will inevitably affect and change social relations, including production relations, and changes in social relations will bring about changes in legal relations. Some of these changes can be absorbed through the interpretation and continuation of existing legal rules, while others require innovation at the legislative level. Meanwhile, just like electricity, communication, and previous computer technologies, artificial intelligence, as a universal technology, is also changing the working form of the legal profession itself. Most of the existing discussions have focused on "substitution", which means whether artificial intelligence will replace the work of lawyers, corporate lawyers, judges, and prosecutors. But the common situation in reality and the foreseeable future is "assisting", "empowering", and "strengthening". Whether it's thinking about how to regulate the application and development of artificial intelligence technology through laws (algorithmic laws), or discussing how to empower the formulation, interpretation, and implementation of laws through artificial intelligence (algorithmic laws), a fundamental fact we need to face is the ubiquitous human-computer interaction phenomenon in the digital society. On the one hand, the law needs to protect human dignity and rights in human-computer interaction scenarios, prevent various small programs and e-commerce platforms from excessively collecting and abusing personal information, and support human subjectivity in the online world manipulated by algorithms; On the other hand, the operational scenarios of the law itself are becoming increasingly digital and intelligent, and human-computer interaction has become a fundamental feature of legal professionals' daily work. The concepts of "wisdom" in smart legislation, smart government, smart courts, smart prosecutors, and smart public security do not refer to human intelligence, but to artificial intelligence. It can be said that human-computer interaction is the main form of current and even foreseeable future legal work; Artificial intelligence will not replace legal professionals, but it will change the way legal professionals work; Machines will not replace people, but those who can use machines will replace those who cannot; It is very important for legal professionals to use technology to enhance their professional abilities.

This article takes human-computer interaction, a common phenomenon in the digital society, as the starting point to analyze how to develop human centered artificial intelligence (Human-centered AI) through appropriate rules and institutional design in the legal and algorithmic dimensions of algorithms, so as to make technology serve the beautiful and good life of humanity.

1、 The Concept and Types of Artificial Intelligence

Starting from the concept of artificial intelligence is to highlight the basic clues of this article, namely human-computer interaction and the principle of human-centered artificial intelligence design. There is no unified definition for artificial intelligence. There is a popular joke among software engineers: "Artificial intelligence refers to something that computers cannot do yet. If it does, we won't call it artificial intelligence, but computer science." This joke actually reveals a very professional judgment: artificial intelligence is actually the profession we used to call computer science, which is the cutting-edge part of the profession aimed at breaking existing boundaries, Enable computers to do things that were previously impossible. A widely influential textbook in the field of artificial intelligence summarizes four ways to define artificial intelligence: machines that think like humans; A machine that acts like a human; A machine for rational thinking; A machine of rational action. This classification can basically summarize the commonly used definitions of artificial intelligence (see Table 1).

Table 1

The concept of artificial intelligence first appeared in the proposal of the Dartmouth Conference. In August 1955, mathematician John McCarthy, computer and cognitive scientist Marvin Minsky, IBM system designer Nathaniel Rochester, and information theory founder Claude Shannon discussed holding a summer seminar the following year. The main purpose of the seminar is clearly stated in the conference planning book: We propose to hold a two-month, ten person seminar at Dartmouth College in Hanover City, New Hampshire during the summer of 1956. This seminar will be based on the assumption that every aspect of learning and other aspects of intelligence can be described so accurately in principle, that we can create a machine to simulate it. We will try to discover how to make machines use language Propose abstract propositions and concepts, solve certain problems currently left for humans to solve, and achieve self-improvement. We believe that as long as a group of carefully selected scientists work together for a summer, we can make one or more significant progress in these areas In the conference invitation letter, the concept of "Artificial Intelligence" was officially born. From the beginning, it can be seen that artificial intelligence was a career that a group of like-minded scientists and engineers intended to pursue, with the aim of enabling machines to learn and complete tasks that humans need to use intelligence to complete. It is an innovation of traditional programming models. The core technology of artificial intelligence is machine learning, and its commercial application is called predictive analytics. But this career was not smooth sailing, and it went through several "cold winters" in the middle. It was only in recent years that it began to receive attention from the general public outside of the small circle and was considered the core technology of the "Fourth Industrial Revolution", with the potential to completely change the survival situation of humanity.

The reason for the sudden eruption of the long-dormant volcano of artificial intelligence is the combined force of economic and technological changes. On the one hand, the internet platform economy has become the mainstream of the contemporary economy, with Google, Baidu, Amazon, Alibaba, Facebook, Tencent replacing manufacturing giants such as General Electric as the new era's commercial powerhouse, with a large number of transactions completed online rather than offline. From wearable devices to household appliances, they have become intelligent and networked, elevating the Internet to the Internet of Things. People's online and offline lives are increasingly integrated, and more and more people are becoming "quantitative selves" measured and supervised by sensors. They generate a large amount of data for analysis and productization every moment, becoming the tracking and analysis object of "ubiquitous computing" (ubiquitous computing or ubiquitous computing). On the other hand, Gordon Moore, one of the founders of Intel, proposed the law that the number of transistors that can be accommodated on integrated circuits doubles every 18-24 months. This law not only applies, but also reflects computer power and storage capabilities, allowing massive amounts of data to be stored and processed at low costs, providing abundant resources for machine learning. Big data is the fuel of artificial intelligence, and application programming interfaces (APIs) are the engine of artificial intelligence. Coupled with the huge profit opportunities of commercial applications, they jointly promote the vigorous development of artificial intelligence.

Compared to previous "non intelligent" tools, the biggest feature of artificial intelligence technology is its "learning" ability. Learning has a special meaning here. In the words of Sima He, one of the theoretical founders of artificial intelligence technology and the winner of the Turing Prize and Nobel Prize in Economics, "learning is any permanent change in the ability of a system to adapt to the environment, whether large or small." The materials of artificial intelligence learning are human behavior data and content data such as text, audio and video. Through learning, It can analyze and predict human behavior (analytical artificial intelligence), as well as generate new content (generative artificial intelligence). That is to say, it can enhance its ability to interact with humans, discover hidden paradigms and patterns in massive amounts of human behavior and content data, and induce human behavior based on this, generating content that people may mistakenly think is created by humans. This "live" or "intelligent" tool is something that humans have never used before, and it has also turned human-computer interaction into true "interaction". It is precisely because of this characteristic that it becomes very important to regulate the application of artificial intelligence technology from the perspective of human-computer interaction interface design, process governance, and result control. Unfortunately, the existing legal rules related to digital technology, including artificial intelligence, still consider people and objects (tools) separately, with some rules starting from the impact of technology on people and focusing on protecting human rights (such as privacy and personal information rights); Another part emphasizes human control over technology (with dedicated personnel responsible for rules in platform responsibility), while lacking consideration and regulation of human-computer interaction as a systematic existence.

Artificial intelligence technology and its commercial applications involve three major elements: computing power, algorithms, and data, corresponding to the physical layer, logical layer (or software layer), and content layer in network architecture. China has currently formulated legal norms for each element. For example, the Cybersecurity Law on Computing Power and the Regulations on the Security Protection of Key Information Infrastructure, the Regulations on the Management of Deep Synthesis of Internet Information Services, the Regulations on the Recommendation of Internet Information Service Algorithms, and the currently being drafted Measures for the Management of Generative Artificial Intelligence Services, as well as the relevant provisions of the Civil Code's personality rights system for data (content) Personal Information Protection Law, Data Security Law, and Regulations on Ecological Governance of Network Information Content. Among them, only the "Regulations on the Ecological Governance of Network Information Content" go beyond the thinking framework of "human object" dichotomy, and propose requirements for the design of human-computer interaction interfaces from the perspective of ecology and architecture. For example, Article 9 stipulates that "a network information content service platform shall establish a mechanism for ecological governance of network information content, formulate detailed rules for ecological governance of network information content on this platform, and improve systems such as user registration, account management, information release review, post comment review, page ecological management, real-time inspection, emergency response, and disposal of network rumors and black industry chain information

After sorting out the existing normative documents, we can find that there are relatively clear legal definitions for the concept of artificial intelligence and the types of artificial intelligence (machine learning algorithms) (see Table 2).