The Copyright Risk Of Machine Learning In The Era Of Artificial Intelligence And Its Solution

The application of machines has liberated human hands,while machine learning has great potential in liberating human brain functions.Different from the previous instrumental approach to creation,machine learning relies on big data and can independently capture tens of thousands of data and information.Regardless of how the data information is identified as a work,if the information that constitutes a work is licensed,then the organization(enterprise)engaged in the development of machine learning technology will be overwhelmed and will seriously hinder the development and application of technology.At the same time,if machine learning makes reasonable use of the data of works,it is very easy for artificial intelligence enterprises to abuse the causes of rights restriction,but harms the interests of authors in a relatively weak position.The use of machine learning works falls into the dilemma of copyright infringement,which casts a shadow over the development of artificial intelligence technology in China.The problem of copyright infringement caused by machine learning is worth thinking deeply.Therefore,this paper USES comparative analysis,empirical analysis and value analysis to explore the copyright risk and resolution mechanism in the development of machine learning technology.Besides the introduction and conclusion,the article is generally divided into five parts:The first part is an overview of machine learning techniques.This part firstly elaborates the theory of machine learning,on the one hand,starting from the concept of machine learning,to perspective of the relationship between artificial intelligence and machine learning,analysis accurate positioning in the field of machine learning in artificial intelligence,depending on the annotation of training data for machine learning classification,with clear study object of this article.On the other hand,starting from the process of machine learning "creation",according to the different tasks of each stage of machine learning "creation",it is disassembled into three stages of data preparation,data utilization and output results,and combined with the infrastructure of machine learning system,it is briefly sorted out the root of contradiction between it and the existing copyright system.The second part is the infringement analysis of machine learning technology in each application stage.First of all,in the data preparation stage,machine learning may be exposed to infringement risks through various data collection methods such as violating the crawler protocol,grabbing database works materials privately,enforcing user license agreement and avoiding technical protection measures.Secondly,in the data utilization stage,the behavior process of machine learning can be broken down into text and data mining behavior and computer data analysis behavior,and there are still different degrees of infringement possibilities.The third part discusses the adjustment of copyright infrastructure under the impact of machine learning technology.First of all,the author makes a retrospective analysis of the legitimacy foundation and rights standard of copyright law respectively to respond to the question of how to construct the infrastructure of traditional copyright law.Secondly,under the background of the increasing uncertainty of copyright law and the imbalance between private rights and public interests,this paper explores the influence of machine learning technology on the internal mechanism of copyright law from the perspective of behavioral science and game theory.Finally,looking into the future,the new pattern of interest balance of copyright law should strike a balance between lenient and strict protection rights holders,and should pay more attention to responding to the demand of technology development for interest sharing.The fourth part focuses on the machine learning in the data preparation stage of the infringement risk to provide a mechanism.From the user’s point of view,the traditional copyright authorization model of "authorization before use" has been somewhat inadequate to meet the needs of new Internet technology subjects to utilize and disseminate works under different business models.Therefore,a flexible copyright authorization mechanism is proposed.For the obligee,the legitimacy of technical protection measures has gradually become the focus of justice.In this paper,the author compares the provisions of the copyright legislation of the United States and the European Union on restrictive technical measures in order to provide a reference for the copyright legislation system of China’s technical protection measures.The fifth part focuses on the mechanism of resolving copyright infringement risk in machine learning data analysis.The first is the legislative evaluation of the mode of copyright exception protection outside China.The United States,the European Union,Japan and other developed countries respectively consider their own legislation tradition of copyright law system and actively carry out legislative or judicial beneficial attempts.It provides an example for the research of copyright exception mode in the stage of machine learning data analysis in China.The second is the legislative construction of the machine learning copyright exception mode in China.Based on the model of fair use rules,Chinese scholars have provided a variety of alternatives to prove the validity of machine learning analysis and use of works.Therefore,on the basis of analyzing the above alternatives one by one,this paper proposes to design specific copyright exception clauses for data analysis behaviors of machine learning according to the Berne Convention and the "three-step testing method" stipulated in TRIPs,so as to clarify the legality of machine learning’s utilization of works.