알고리즘의 인지 차이가 중국 인공 지능법률 규제에 끼치는 부정적인 영향 및 개선방안 연구

Since 2020, artificial intelligence norms raised by international organizations, the EU, and the United States have adopted a position that respects the definition of algorithms of a technical nature. Under these circumstances, research guided by actual law should not be overly expanded based on reality. Recently, differences in perceptions such as beliefs, negative prejudices, and crowd psychology have seriously distorted people’s objective perceptions of algorithms. Only by excluding any interference can the nature of artificial intelligence algorithms and effective legal regulations based on them be clarified. The difference in perception of algorithms appears in the form of data, computing power, artificial intelligence, and models. Algorithms cannot fundamentally give the sophisticated computing characteristics of artificial intelligence if the difference between algorithms and operations is discussed. Algorithms in a simple mathematical sense have no direct property value and are not exclusively protected by the law, even in principle. Article 25 of China’s Patent Act stipulates that the rules and methods of intellectual activity do not grant patent rights, but algorithms belong to the rules and methods of intellectual activity that solve specific problems. Naturally, patent rights can be granted even if the technology that solves technological problems and generates technological effects meets the needs of novelty, creativity, and practicality. However, the key is the grant of rights to the technology, not the algorithm. On the other hand, if algorithmic perception is left unbiased, AI law regulation must discuss algorithm regulation because serious problems arise in its object, basis, purpose, and level. The most reliable regulatory route is in the data and not in the algorithm because the regulation of algorithms is visualized through the enactment of law and response through data laws. However, the development of artificial intelligence must continue through the collection, preservation, and processing of data, such as device learning, 先 experience knowledge of data, and inadequate data processing. Algorithm management becomes a kind of search tool, and if the search shows a bias far from social norms, it is not appropriate and can lead to situations where overconsensus is difficult. On the other hand, the scope of the adjustment type of the “Internet Information Service Algorithm Recommendation Management Regulations” includes “generated synthesis, individualized exports, ordered selection, search filtration, and arrangement crystals.” What is clear is that the nature of the “Internet Information Service Algorithm Recommendation Management Regulations” is still based on content management. In particular, it is necessary to prevent related services from affecting normal broadcasting, markets, and social order, to be harmful to the safety of consciousness, social fairness, and equality, and to avoid causing damage to the legitimate rights and interests of the Chinese people. Registration and public announcement problems include the quality of public announcement information, algorithm classification framework, risk description method, and regulatory target setting. A way to improve algorithm management based on accurate algorithm recognition is to establish stronger disclosure content and quality demands. The disclosure content should be renamed to general information on an artificial intelligence system or service system. In addition, in order to reduce the difficulties caused by the lack of algorithm classification and grade classification, it is necessary to prevent the algorithm from intervening excessively in the dominant object. A clearer solution is to look at the relationship between norms and development. There is no need to unnecessarily subdivide and define the type of algorithm, and on the contrary, the risk rating of the service should be considered as the main consideration requirement. For services with low or intermediate risk, it is desirable to deregulate and focus management on high-risk services with a very large impact on public interests and personal rights. In addition, by referring to how Europe determines the risk level in the application area, it is possible to further subdivide the risk classification description for artificial intelligence services.