Algorithm Regulation In Deep Learning Lifecycle

At present,with the rapid development of science and technology,algorithm has brought earth-shaking changes from entering the public’s vision to integrating into daily life,from simple and single structure to rich and diverse types,but also brought challenges to the existing legal and regulatory system.In the tide of artificial intelligence,deep learning algorithm is one of the algorithms with the most cutting-edge technology and the largest development space.It created ChatGPT,Siri and many other software,which is deeply engaged in computer vision,speech recognition,word processing and other fields.However,due to its different characteristics from traditional algorithms,it also brings impact to the existing laws and regulations on algorithm supervision,such as algorithm interpretation power and transparency issues,privacy risks,security issues,intellectual property issues and so on.This paper focuses on the deep learning algorithm itself and its unique life cycle of ’input’ training,output and cycle’,analyzes the four links one by one,identifies the problems and proposes solutions.For the deep learning algorithm itself,the difficulty in implementing the interpretation right of the algorithm has always troubled users.The author believes that the classification management of the algorithm and the strengthening of administrative management are expected to solve the problems.In the life cycle of deep learning algorithms,different legal problems are brought about in the process of data collection,model training,content output and recycling.Therefore,the concept of collaborative processing of legal governance and scientific and technological governance should be adhered to to write a Chinese plan for deep learning algorithm governance.