My opinion on the development and application of MGPT

In order to cope with the challenges and inherent limitations in the development of medical GPT technology, the author suggests a comprehensive and in-depth innovation strategy, which requires a refined reshaping of every link from data acquisition to system operation and maintenance. Data collection is no longer just a quantitative accumulation but a qualitative leap, which means carefully selecting from a vast amount of medical information to ensure that each piece of data is highly representative, accurate, and usable. This process requires not only the support of advanced technical means, but also the in-depth participation of medical experts to achieve accurate data screening and value mining. In the selection of the pedestal model, the traditional simple question and answer framework should be abandoned, and the possibility of building an expert digital human doppelganger should be explored. This transformation allows patients to receive more personalized and professional medical consultation services as if they were directly facing experienced doctors, which greatly improves the interactive experience and trust. At the same time, in order to ensure the security and accuracy of medical information, it is recommended to apply an AI-based intelligent quality control mechanism to replace the blind reliance in the past, and strictly control the quality through a combination of automatic review by algorithm and manual review. In addition, the training, evaluation and optimization of models should also pay more attention to the integration of practical experience. On the basis of evidence-based medicine, it advocates the integration of the clinical wisdom and experience of big doctors into the model, so that medical GPT technology can not only provide patients with more accurate and individualized diagnosis and treatment suggestions based on the latest scientific research results, but also combine with clinical practice. In short, it is necessary to realize the four major transformations from data cleaning to selection, from simple consultation to expert clone, from blind trust to quality control, and from simple evidence-based to combined with the experience of doctors.

Key Words: artificial intelligence; generative pretrained transformer; medical generative pretrained transformer; natural language processing; open evidence