2503-人工智能机器学习与监管_自动驾驶汽车案例_57页_3mb

CPB Corporate Partnership Board
Summary and Conclusions
AI, Machine Learning and Regulation: The Case of Automated Vehicles
International Transport Forum (ITF)

Automated vehicles (AVs) leveraging AI and Machine Learning (ML) promise enhanced safety and accessibility in transport but introduce challenges such as data quality issues, verification difficulties, societal impacts, and ethical concerns. The Safe System approach prioritizes safety by assuming mistakes are inevitable and ensuring they do not lead to serious injuries or deaths. A key issue is verifying AV safety, as traditional methods must evolve to account for AI's dynamic behavior and potential biases in data.

Policy recommendations include focusing regulations on fundamental principles like safety, explainability, and human oversight. Authorities should mandate reporting of safety-relevant incidents, develop diverse test scenarios, support machine-readable infrastructure, and promote explainability to build public trust. The entire AI lifecycle—from data collection to deployment—must be addressed to ensure trustworthy AVs that benefit society.