This national survey of academic critical care physicians in China compared two cohorts surveyed a year apart, before and after the public release of the DeepSeek large language model. Self-reported use of generative AI jumped from 65% to 94%, yet participation in formal training stayed flat at about 13%. Structured training was strongly associated with higher self-reported competence, and the authors argue the focus must shift from simply adopting AI to teaching clinicians how to integrate it safely.
