Summary of Key Points
The development of large medical AI models relies heavily on the input of medical expertise from doctors. However, those doctors who participate in AI training, data annotation, and evaluation often face challenges such as increased workloads, reduced salaries, limited model improvements, and diagnostic approaches that deviate from the essence of healthcare, leading many to leave their jobs. This situation highlights the real dilemmas within the medical AI industry: is AI a tool to assist doctors or a potential competitor? How can doctors find their place in this emerging technology?
1. What Do Doctors Do for AI?
Doctors play a crucial role in training AI by annotating medical records and evaluating its performance:
- Zhong Sheng (Annotator): Extracts key information such as diagnoses, medications, and procedures from patient records to label the data for AI analysis. For example, for lung cancer patients, details like the size of tumors and genetic mutations must be accurately labeled. This task can take over 20 minutes for a patient with a long medical history.
- Zhou Yan (Evaluator): Assesses AI systems developed by large companies to determine their accuracy in answering user questions and identifying potential risks. For instance, if an AI system simply recommends going to the hospital without providing emergency treatment advice or comfort, it receives a low score. He also provides feedback to help improve the models.
- Lin Jing (Part-time Expert): Reviews skin disease images to diagnose possible conditions but finds that diagnostic standards need to be standardized to avoid unnecessary rework.
These tasks may seem like teaching roles, but in reality, doctors are more like data processors—repeating mundane and mechanical tasks while having to adapt to the pace of AI development.
2. Why Are Doctors Leaving the AI Field?
Doctors leaving the AI field face various compelling reasons:
- Zhong Sheng: Went from a monthly salary of 20,000 yuan to a reduced income due to increased workloads and company restructuring. The company expanded tasks (from 150 patients to 300) while cutting salaries, forcing him to switch to a private hospital as a doctor.
- Zhou Yan: Fears being replaced by AI, especially after observing rapid model improvements that could lead to his job loss. He decided to transition to a medical researcher to stay involved in the entire AI development process.
- Lin Jing: Disagreed with company policies that prioritized efficiency over comprehensive diagnoses, which conflicted with her professional standards as a top-tier doctor.
Their departures reflect the tension between the practical realities of the AI industry and doctors' career aspirations—either being treated as mere tools or worrying about being replaced by technology.
3. What Is the Current Level of Medical AI?
From doctors' perspectives, current AI systems are still quite limited:
- Rigid Guidance: Models provide fixed treatment plans based on predefined guidelines, lacking flexibility.
- Lack of Precision: Some AI products neglect rare diseases for efficiency, compromising diagnostic accuracy.
- Slow Progress: After three years of working with AI, there has been little improvement in the models' capabilities.
In summary, medical AI is currently more like a reference tool that can look up basic medical information but still struggles to handle complex cases and rare diseases effectively.
4. Doctors' Dilemmas: Is AI a Help or a Threat?
Doctors have mixed feelings about AI:
- Zhong Sheng: Initially hopeful about using AI to transition into coding, he became disillusioned and returned to clinical practice.
- Zhou Yan: Believes in the potential of medical + AI but realizes the need for skill enhancement to avoid being obsolete.
- Lin Jing: Is confused about the direction of AI development, questioning whether it should focus on profit or patient care.
These dilemmas reflect broader issues within the industry: AI must balance efficiency and quality, and integrate technology with clinical practice effectively to gain doctors' trust and collaboration.
5. The Impact of Competition in the Medical AI Industry
Small companies struggle in this highly competitive field, often leading to increased workloads and reduced employee compensation:
- Zhong Sheng’s Company: Unable to compete with larger firms, it cut costs by increasing staff workload and reducing salaries, eventually resulting in job losses.
- Zhou Yan’s Experience: High-paying evaluation positions at large companies require meeting unrealistic performance standards, turning employees into data processors.
The struggle of small companies and the forced departures of their employees highlight the harsh realities of the medical AI industry, where not all participants benefit from its growth.
Conclusion
Medical AI cannot thrive without doctors. However, doctors face challenges such as being treated as tools or facing job displacement. For AI to truly become a valuable asset, it must address issues related to efficiency and ethical considerations, ensuring that it enhances healthcare rather than replacing doctors’ expertise and empathy. Doctors, in turn, need to find their role in this evolving landscape—either by designing AI systems or mastering how to use them effectively—to avoid being left behind by the times.