论人工智能透明度原则的法治化实现

On the Realization of the Principle of Transparency of Artificial Intelligence by Law

随人工智能应用范围不断扩大,透明度原则成为破解“算法黑箱”难题的钥匙。透明度原则包括形式透明和实质透明。形式透明是对人工智能基础信息的披露,使人工智能的部署使用处于非秘密状态。实质透明与人工智能的可解释性息息相关,强调对披露信息以可被理解的方式进行有意义解释,从而打破知识壁垒,使相关信息实现真正可知。透明度原则应在法治化实现上进行精细考量:在考虑现有技术条件的基础上,有必要根据监管机关和社会公众的能力不同有区别地对披露和解释的信息内容提出义务要求,并针对不同人工智能的风险程度采取分级分类分场景的执行标准,从保护用户与公众知情权行使、确保国家监管权落实两方面推动透明度原则实现。

With the continuous expansion of the application of artificial intelligence,the principle of transparency has become the key to solving the problem of “algorithm black box”.The principle of transparency includes transparency in form and transparency in substance.Formal transparency refers to the disclosure of AI-related information so that the information is non-secret.Substantive transparency is closely related to the explainability of artificial intelligence,emphasizing the interpretation of disclosed information in a way that can be understood,to break the knowledge barrier and make relevant information truly knowable.The principle of transparency should be carefully considered in the realization of the rule of law:On the basis of considering the existing technical conditions,it is necessary to put forward requirements for the disclosure and interpretation of information content according to the different capabilities of the regulatory authorities and the public,and adopt the implementation standards according to the risk degree,classification and scenario of different artificial intelligence,and improve the regulatory system from the path of protecting the user and the public's right to know and ensuring national supervision.