Trusted Artificial Intelligence: Technique Requirements and Best Practices
Tao Zhang, Yi Qin, Qiang Li · 2021
The development and adoption of artificial intelligence (AI) have brought numerous problems and potential threats, such as information cocoons, algorithmic collusion, algorithmic bias and privacy, issues. To combat these issues, governments all over the world, international organizations and giant tech companies all have taken actions and reached an agreement that AI should be trusted. However, there is no uniform definition on trusted AI. In the paper, we make a survey on related works on artificial intelligence principles and ethical guidelines. Then analyze the ethical foundations of trusted AI and give a definition on trusted AI. We propose specific requirements of trusted AI from a technical perspective and give explanations on these technique requirements. Finally, best practices are recommended to achieve trusted AI and promote the responsible use of AI.