医院布局大模型很热闹，缘何还难以真正落地

Core Viewpoint - The article discusses the challenges and advancements in the integration of AI, specifically the DeepSeek model, in hospitals, highlighting the cautious optimism of medical professionals regarding AI's potential to improve healthcare delivery and operational efficiency [1][2]. Group 1: AI Implementation in Hospitals - Many hospitals are actively deploying DeepSeek, but the expected improvements in healthcare delivery are still below expectations, with some medical staff either not using it or struggling to adapt [1]. - The Shanghai Oriental Hospital, under the leadership of Dr. Duan Tao, is collaborating with the Institute of Software, Chinese Academy of Sciences, to develop the Medgo AI model, which has received an A-level recommendation from Shanghai's medical AI application testing center [2]. - The integration of AI in hospitals is seen as a gradual process, requiring iterative improvements and cautious adoption in specific departments before broader implementation [2]. Group 2: Cost and Accessibility of AI - The cost of developing AI applications for hospitals has significantly decreased, from several million yuan to around 5 million yuan, making it more accessible for various healthcare institutions [3]. - Hospitals are encouraged to evaluate their individual needs regarding AI adoption rather than rushing into widespread implementation [3]. Group 3: Challenges in AI Application - Current explorations of AI in hospitals focus on enhancing patient experience, medical services, and hospital management, with the most significant challenges arising in the area of medical service applications [4]. - The accuracy of AI models is heavily dependent on the quality of the data fed into them, with inconsistencies in medical terminology posing challenges for data standardization [4]. - There is a consensus that while AI can improve efficiency, especially in administrative tasks, its application in clinical settings remains limited due to high medical standards, sensitive data issues, and a shortage of cross-disciplinary talent [5]. Group 4: Future Outlook - Despite current limitations, there is confidence in the ongoing evolution of AI technology, with expectations that it will eventually penetrate deeper into clinical diagnosis and treatment [5].