Preview of Chengdu Big Model｜Dialogue with Wang Shirui of Yilian: Medical Big Model and Doctors Are Not Substitutes for Each Other
2024-08-14
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
Preview Chengdu Large ModelThe development of artificial intelligence is in full swing, integrating with thousands of industries and generating new technological power. As the key to the competition of artificial intelligence, big models will enter the first year of monetization this year.Red Star Capital noted that in this industry competition, the compound growth rate of Chengdu's artificial intelligence industry has exceeded 40% in the past three years, and its "competitiveness of artificial intelligence technology industry" has ranked sixth in the country. Sichuan Province has even listed artificial intelligence as the "No. 1 Innovation Project" this year.Recently, Red Star Capital interviewed several AI large model manufacturers in Chengdu and discussed with them the "Chengdu power" in the artificial intelligence trend.
Driven by the wave of digitalization, artificial intelligence is gradually penetrating into the entire medical industry. Last year, Yilian, a pioneer in domestic Internet hospitals, also launched China's first large-model-driven AI doctor - MedGPT.
Red Star Capital learned that the original intention of developing MedGPT stems from a deep insight into the needs of the medical field in the digital economy era. Yilian hopes to achieve intelligent diagnosis and treatment of the entire process of disease prevention, diagnosis, treatment and rehabilitation through MedGPT, thereby improving the efficiency and quality of medical services and optimizing patients' medical experience.
The road to innovation is not always smooth. How does MedLink overcome challenges such as the inability of AI to have continuous and free conversations with patients and the improvement of medical reasoning and diagnostic capabilities? How to bring into play the actual diagnostic and treatment value of MedGPT in real medical scenarios? Red Star Capital Bureau interviewed Wang Shirui, the founder of MedLink, and discussed in depth the story behind the development of MedGPT, the challenges it faces, and its far-reaching impact on the future of intelligent medical services.
(Photo provided by interviewee Wang Shirui)
Red Star Capital Bureau: Why did Yilian want to be China's first large-scale model-driven AI doctor "MedGPT"? What are the biggest challenges faced during the research and development process?
Wang Shirui:With the continuous penetration of AI technology in various industries, the digital economy has become an important engine leading industrial development. In the medical field, the application of AI is particularly important because it is related to the national economy and people's livelihood. Yilian hopes to use MedGPT to play the actual diagnosis and treatment value in real medical scenarios and realize the full process of intelligent diagnosis and treatment capabilities from disease prevention, diagnosis, treatment to rehabilitation.
The main challenges are: the inability of AI to have continuous and free conversations with real patients; how to support multimodal input and output in diagnosis and treatment scenarios; and how to enable MedGPT to have strong medical reasoning and decision-making capabilities.
To address these challenges, MedLink uses the Transformer architecture to enable MedGPT to guide patients to collect sufficient diagnostic decision factors through multiple rounds of consultations to ensure the accuracy of the diagnosis. At the same time, MedGPT integrates multiple medical testing modalities to enable patients to be prescribed necessary medical examination items after the consultation to further clarify their condition. Finally, through deep learning and training on massive medical data, MedGPT can accurately understand the patient's description of the condition, provide personalized diagnosis and treatment recommendations, and assist doctors in making decisions.
Red Star Capital Bureau: When it comes to medical issues, the accuracy of general large language models has natural defects, and may even lead to wrong judgments due to the feeding of erroneous data. How does MedGPT solve this problem?
Wang Shirui:First, MedGPT will guide patients to collect enough diagnostic decision factors through multiple rounds of consultation to ensure the accuracy of the diagnosis. At the same time, Yilian organized a large number of doctors and medical experts to conduct secondary training on MedGPT. Yilian has established a "medical expert system" internally, and has invested more than 100 doctors in manual feedback supervision and fine-tuning training to evaluate the accuracy, efficiency, comprehensiveness, error rate, risk warning level, etc. of the consultation. MedGPT has also launched the world's largest medical AI clinical application trial in cooperation with hospitals such as West China Hospital of Sichuan University, optimizing the model through clinical trial data to improve the accuracy and efficiency of diagnosis and treatment.
Red Star Capital Bureau: What is the current application and commercialization status of MedGPT?
Wang Shirui:Currently, there are representative applications on both the B-end and the C-end. For the C-end, Yilian launched the Jiaxiaoyi Express Edition product. For different groups on the B-end, Yilian provides customized AI solutions around doctor education, smart hospital construction, health management center, insurance, etc., with a transaction amount of tens of millions of yuan. In the future, with the cooperation of some public hospitals, Omron, Peking University Resources Holdings, Laobaixing Pharmacy and other industry leading brands with Yilian, the subsequent market will have further room for improvement.
Red Star Capital Bureau: What are Yilian’s expectations for MedGPT? Has it been met? Are there any areas that can be improved?
Wang Shirui:Yilian hopes that MedGPT can play a practical role in diagnosis and treatment in real medical scenarios, and realize the whole process of intelligent diagnosis and treatment from disease prevention, diagnosis, treatment to rehabilitation. At the same time, Yilian expects MedGPT to continue to assist doctors through its technological innovation, improve diagnosis and treatment efficiency, and optimize patients' medical experience.
Currently, MedGPT has met these expectations to a certain extent. For example, in recent real-person tests, MedGPT's diagnostic hit rate reached 80%, which is much higher than the average diagnostic hit rate of about 35% of existing medical models, and is close to or even exceeds the diagnostic level of real doctors. In addition, MedGPT has been able to cover the first diagnosis capabilities of nearly 3,000 diseases, covering more than 80% of adult diseases and more than 90% of pediatric diseases aged 0-12 years.
However, MedGPT still has room for improvement. For example, further improving the ability to communicate with patients in natural language, enhancing the ability to handle complex and systematic problems, and improving the accuracy and efficiency of diagnosis and treatment are still the key to the future development of MedGPT.
Red Star Capital Bureau: There are many large medical models at present. What is the core competitiveness of MedGPT?
Wang Shirui:Medlink has been focusing on the serious medical field for 10 years and has accumulated a lot of experience and high-quality data. This means that it has industry-leading experience in medical links such as prevention, diagnosis, treatment and rehabilitation, and can conduct more targeted in-depth learning and optimization when developing MedGPT. In addition, Medlink has cooperated with nearly 40 experts at the chairman level of the Chinese Medical Association to build core product strength through the highest quality medical data. That is, "I have what others don't have, and I am better than others."
Red Star Capital Bureau: Do you think the big medical model will replace doctors?
Wang Shirui:Medical big models will not replace doctors because they are essentially auxiliary tools, not replacements.
In terms of auxiliary diagnosis, efficiency improvement, data analysis, knowledge updating, and personalized patient treatment, the medical big model and doctors can form a complementary relationship and jointly improve the quality and efficiency of medical services rather than replace each other.
Red Star Capital Bureau: In the entire international medical big model track, what are the advantages and challenges of domestic medical big models?
Wang Shirui:In the entire international medical big model track, domestic medical big models have two major advantages. One is the data scale. The huge population base in China provides rich data resources for big model training. The second is the policy environment. The support of national policies provides a good environment for the research and development and application of medical big models. The challenges we need to overcome are first, the training of interdisciplinary talents. China needs more interdisciplinary talents, especially at the junction of medical and AI fields. Second, in terms of algorithms, computing power and model architecture, there may still be a certain gap between us and the international advanced level.
Red Star Capital Bureau: What challenges or bottlenecks does the medical large model industry currently face?
Wang Shirui:The first is data security and privacy protection. Because the medical industry has strict requirements for data security and privacy protection, big models need to ensure data security and compliance when processing sensitive medical information, which is a major challenge. The second is commercial implementation and profit model. Although medical big models have made progress in technology, how to build a sustainable business model and solve the problem of input-output efficiency is one of the main challenges currently facing medical AI companies. The third is supervision and compliance issues. Medical big models need to face strict supervision and compliance requirements in clinical applications to ensure the safety and effectiveness of their applications.
Red Star Capital Bureau: Yilian was born in Chengdu. What empowerment has Chengdu’s big model industry chain brought to Yilian’s research and development of medical big models?
Wang Shirui:The rapid development of Yilian is inseparable from the support of Chengdu. The release of relevant documents such as "Implementation Rules of Several Policy Measures for Further Promoting the High-Quality Development of the Artificial Intelligence Industry in Chengdu" has provided strong support for the development of the artificial intelligence industry.
For Yilian's research and development of medical big models, Chengdu's big model industry chain provides multi-faceted empowerment. First, Chengdu's support policies and action plans provide a good policy environment for Yilian. Secondly, Chengdu's existing big model development foundation and industry chain provide support for Yilian in terms of technology, data, and talents. In addition, Chengdu's big model industry ecosystem is also gradually maturing, which provides a platform for cooperation and communication for companies such as Yilian, and helps promote the development and application of medical big models.
Red Star News reporter Zhang Luxi
Editor: Yang Cheng
(Download Red Star News and get a reward for reporting!)
