The rapid rise of AI large language model(LLM)technologies,represented by ChatGPT,is revolutionizing content pro-duction and intelligent technology paradigms.However,issues such as hallucinations and false content have led to a crisis of trust,with the technology even facing resistance and bans.Despite the industry's active technical practices in the field of trustworthy AI,public trust in AI has not significantly improved.Therefore,to address the trust issue,it is not only necessary to clarify the relationship be-tween trust and trustworthiness but also to consider the technological essence of LLMs.As an epistemic technology,trust in LLMs should be epistemic trust,which includes the dynamic interaction of technical trust and interpersonal trust,and is based on reasonable trust established through effective supervision.Accordingly,a trust construction route for LLMs is proposed,which mainly includes a trust element system with being explainable at its core,a trust subject and environment with a government-led AI governance system as the basis and the collaboration of multi-subjects,and the trust cognition that cultivates a correct trust perspective among people.
artificial intelligencelarge language modelstrusttrustworthy
NETL
NSTL
万方数据
