（Translated By YU.Y.C.）On July 12,Professor Wong Tien Yin, a well-known expert in ophthalmology of Tsinghua University and Beijing Tsinghua Changgung Hospital,Professor Ma Weiying and Professor Li Jingshan from Tsinghua University jointly published an opinion article in the New England Journal of Medicine · Artificial Intelligence (NEJM AI), together with many scholars from National University of Singapore and Shanghai Jiao Tong University. This paper discusses the evolution direction and key technologies of medical artificial intelligence model, proposes the concepts of UMAIand UHAI, and summarizes the key challenges and technology paths.

The article points out that existing medical AI technologies still face several challenges, including: current models have failed to learn to integrate "human intelligence" accumulated based on clinical training; The current model pays more attention to the treatment of disease states, but neglects the evolution process from health to disease. Current models rely on data collected from developed countries and ignore medical problems in developing countries.

Given the above challenges, the paper proposes two new paradigms for medical AI: UMAIand UHAI. The main features of both include the integration of human clinical experience and the integration of out-of-clinical health data. The key technical content of both include health data integration and alignment, human values alignment, medical dark knowledge injection, strong foundation models, etc.

The article said that UMAI and UHAI, as innovative model paradigms, can try to make up for the shortcomings of medical artificial intelligence in transformation and application. By combining tacit knowledge and non-clinical health data, they promise to construct more comprehensive medical AI models to address the many challenges in healthcare.