A project in China to develop autonomous and self-developing virtual healthcare is expected to go public next year.
This confirms it by Yang Liu, professor at the Department of Computer Science and Technology at Tsinghua University and co-research director of the Agent Hospital project. The concept of a virtual hospital developed by scientists from the university The Artificial Intelligence Industry Research Institute (AIR) simulates the actual cycle of the hospital treatment process, from the onset of illness to follow-up. The Institute considers this concept to be the first of its kind in the world. The results of this study were first published in May this year arXiv, Cornell University’s open-access online repository of research articles.
WHY IT’S IMPORTANT
All virtual actors in Agent Hospital, including patients, nurses, and doctors, are generated using a huge language model (LLM). These AI characters will represent real humans when the system is released to the public in the first half of 2025. The public pilot, which will be led by Tairex, an AIR spinoff startup, will begin sometime in the first quarter, Professor Yang said.
For the virtual hospital concept, researchers proposed a design method called MedAgent-Zero that enables artificial intelligence physicians to continuously learn, improve, and enhance accuracy in performing clinical tasks by interacting with patients, reviewing medical literature, and accumulating experiences from managing both successfully , as well as failures. matters.
The study results showed that with this novel method, AI doctors achieved 88%, 95.6% and 77.6% accuracy in examining, diagnosing and treating patients, respectively.
“A doctor can diagnose and treat tens of thousands of patients in a matter of days, which typically takes at least two years for a human doctor,” the researchers also noted.
Meanwhile, the AI physician also demonstrated accuracy of up to 93% in answering a subset of the MedQA dataset – primarily based on a competitive US licensing survey that included questions about major respiratory diseases.
As part of the development of the concept, researchers plan to expand the scope of diseases and expand it to more medical departments. There are currently 42 AI doctors working on the virtual platform in 21 medical departments, including emergency, respiratory and cardiology departments.
They also plan to include more features, including promotions for medical positions, changes in the distribution of diseases over time, and historical patient medical records.
There is also a plan to optimize the selection and implementation of the base LLM. Their research currently uses OpenAI’s ChatGPT models versions 3.5 and 4. “We will be using the latest and most advanced LLM,” Professor Yang said.
A BIGGER TREND
Other research initiatives in China have also developed medical LLM programs for clinical decision support. As part of a project carried out at the Faculty of Medicine at Tongji University, a model called MedGowhich was trained using 6,000 medical textbooks and has since been integrated and used in the affiliated Shanghai East Hospital.
The Institute of Artificial Intelligence of the Chinese Academy of Sciences – one of China’s national research centers – introduced it earlier this year CARES Copilot chatbot based on Meta’s Llama 2 LLM platform that helps doctors make diagnoses and therapies.