摘要
近年来自动驾驶伦理中有一个知名度与认同度都很高的观点:算法迭代会导致自动驾驶责任主体缺失。重审发现,该观点混淆了迭代过程与算法能力。算法供应商虽然难以控制机器学习型算法的迭代过程,却仍能控制最终获得的算法能力的水平,包括感知识别型算法的识别准确率,路径规划型算法与行为决策型算法的合理程度。在健全的算法产品准入与质检制度下,算法供应商对算法产品承担有限的产品责任。
In recent years,there has been a well-known and widely recognized opinion in ethics of autonomous ve⁃hicles:iterative algorithm will lead to the absence of subject of responsibility in autonomous vehicles.It is found that this opinion confuses process of iterative algorithm with algorithmic capabilities.Although it is difficult for algo⁃rithm suppliers to control the process of iteration in machine learning algorithms,they can still control the level of algorithm capabilities eventually obtained,including the accuracy of recognition of perceptual recognition algorithm,the reasonable degree of path planning algorithm and behavior decision algorithm.Under the sound algorithm prod⁃uct access and quality control system,algorithm suppliers are responsible for limited product liability for algorithm products.
作者
李大山
LI Da-shan(School of Marxism,Shanghai University,Shanghai 200444,China)
出处
《科学技术哲学研究》
CSSCI
北大核心
2023年第5期116-121,共6页
Studies in Philosophy of Science and Technology
基金
国家社会科学基金重大项目“负责任的人工智能及其实践的哲学研究”(21&ZD063)
关键词
自动驾驶
算法迭代
责任主体
算法能力
机器学习
autonomous vehicles
iterative algorithm
subject of responsibility
algorithm capabilities
learning ma⁃chines
作者简介
李大山(1992-),男,浙江温州人,哲学博士,上海大学马克思主义学院讲师,研究方向为应用伦理学、元伦理学、认知科学哲学。
