1. Tsinghua University researchers have developed a virtual "Agent Hospital" for training AI doctors without human intervention. The system simulates a hospital environment, including patient interactions, medical procedures, and diagnostic tasks, allowing AI models to learn and improve their medical skills through extensive practice.

2. The Agent Hospital utilizes reinforcement learning, where the AI agents receive rewards or penalties based on their performance in the simulated scenarios, incentivizing them to learn and optimize their decision-making processes. This approach aims to create AI-powered medical assistants that can complement human doctors, enhancing the overall quality of healthcare.

3. The researchers believe that the Agent Hospital can significantly accelerate the development of AI-powered medical solutions, as it provides a safe and controlled environment for testing and refining these systems. By training AI models in this virtual setting, the researchers hope to overcome the challenges of limited access to real-world medical data and the risks associated with direct patient interactions during the early stages of AI development.