信与不信:通用人工智能信任危机的理论反思与机制再造

Believe It or Not:Theoretical Reflection and Mechanism Re-engineering in the Trust Crisis of Generalized Artificial Intelligence

人类是否信任人工智能是现代社会必须直面的一个现实难题,尤其是以ChatGPT为“奇点”的通用人工智能技术的创新与扩张,给人工智能信任的理论供给与制度安排带来了新挑战。在通用人工智能诱发新的信任危机的背后是技术突变在发挥着决定性作用,因而应在“主体符合性—要素流转性—失信可追责性”信任机制的基本框架下,对通用人工智能技术的突变、技术伦理与信任伦理的冲突及其可能引发的信任要素流转等展开深入分析。通用人工智能信任危机受到学习黑箱性与可解释性的对立、自我学习与道德筛选机制的冲突、主体不明晰与责任可追究的矛盾等深层因素的影响,须从技术路径前置、道德路径中置、监管路径后置三个维度再造通用人工智能信任机制,以实现人类智能与人工智能长期的良好互动与有序共融。

ion on the mutation of generalized artificial intelligence technology,the conflict between technology ethics and trust ethics,and the flow of trust elements that may be caused.It is found that the trust crisis of generalized artificial intelligence is affected by the deep theoretical factors,such as the opposition between black box learning and interpretability,the conflict between self-learning and ethical screening mechanism,and the contradiction between unclear subject and accountable responsibility,etc.It is necessary to reconstruct and rebuild the trust mechanism of general artificial intelligence from the three dimensions of the technological path to the front,the ethical path to the middle,and the regulatory path to the back,in order to realize the long-term good interaction and orderly cohesion between human intelligence and artificial intelligence.