Traditional Chinese Medicine Embraces the Age of Big Models
TMTPOST-- Technological innovations have met Traditional Chinese Medicine (TCM), heralding a new chapter where large models—the artificial intelligence products—are increasingly relevant to this ancient practice.
The competitive landscape has drawn in technology giants, innovative TCM companies, research institutions, and even local governments, all eager to stake their claim in the burgeoning field of TCM big models.
Historically, TCM has been a subject of mixed reviews. While there are accolades for its synergy with big models, skepticism persists, with some questioning whether this integration is merely a flash in a pan.
Current applications of TCM big models span drug research, clinical assistance, diagnostic support, and health management. And the advent of AI-driven TCM robots, which mimic traditional practices like pulse diagnosis and prescription writing, has sparked interest in the modern application of TCM. These robots, often dubbed "cyber TCM," reflect the growing convergence of tradition and technology.
Gu Gaosheng, Chairman of TCM AI company TCM Brain, views the integration of TCM with big models as a revolutionary shift in traditional TCM services. He highlights that consumer-facing wellness services, such as those provided in health institutions and pharmacies, are particularly promising. For instance, TCM Brain's "Brain Quick Q&A" model, now in its third version, has significantly improved efficiency and accuracy in TCM practice through advancements in deep learning and big model technology.
Similarly, the "Shuzhi Bencao" model developed by Tasly Pharmaceutical accelerates drug discovery and development by utilizing big models to mine and summarize theoretical evidence from TCM. This model aids in optimizing herbal formulations and predicting the efficacy and safety of new drug candidates, thus enhancing the drug development process.
The development of TCM big models follows two primary technological paths: deep learning and knowledge graphs. These approaches are complementary, with deep learning focusing on understanding and generating language, while knowledge graphs excel in organizing and interpreting complex information.
"Without practical scenarios, there can be no subsequent data feedback; a TCM big model is merely a castle in the air." Many respondents indicated that big models need to be applied in specific scenarios to optimize their performance through practical use.
Li Wenyou, the CEO of Da Jing TCM, emphasizes the importance of integrating both technologies. The "Qihuang Q&A" model of Da Jing TCM leverages a robust knowledge graph developed over eight years and combines it with deep learning to enhance the model's performance and accuracy. This integration demonstrates how big models can revolutionize TCM diagnostics and treatment by providing more precise and contextually relevant insights.
High-quality data is crucial for training effective TCM big models. Key data elements include static sources such as classical texts and clinical guidelines, as well as dynamic clinical data from real-world practice.
However, challenges arise from inconsistent data recording and the private nature of clinical information. To address these challenges, TCM Brain collects data through scalable systems and platforms that aggregate information from thousands of medical institutions. They also utilize their proprietary platform to capture and apply the experiences of leading TCM practitioners.
Data quality directly impacts the performance of big models. Proper data collection, cleaning, and preprocessing are essential to ensure that the models are trained on accurate and comprehensive information. For example, Da Jing TCM has spent eight years developing a standardized terminology dictionary for TCM symptoms and signs, which contributes to the effectiveness of their models.
The development of TCM big models also faces several hurdles. These include the need for interdisciplinary talent, user acceptance, and intellectual property issues. Huang Xinting, the deputy director of the Smart Integration of Chinese and Western Medicine Research Center, stresses the importance of user acceptance, noting that good diagnostic results alone are not enough if the system imposes burdens on users.
Gu points out that intellectual property and industry regulations must evolve to address the unique aspects of TCM. Furthermore, training cross-disciplinary professionals who understand both TCM and AI technology is crucial for advancing the field.