China unveils ChatGPT-like medical chatbot service

The Centre for Artificial Intelligence and Robotics (CAIR) under the Chinese Academy of Sciences (CAS) has developed CARES Copilot 1.0, a chatbot based on Meta's Llama 2 LLM, to aid doctors in medical diagnoses and treatments. Utilising 100 GPUs from Huawei and Nvidia, the chatbot processes diverse data types, including images, text, voice, video, and medical scans like MRI and CT scans. Trained on millions of records, CARES can retrieve information within seconds with up to 95% accuracy, claims CAIR. It boasts a 100K context window and can extract data from over 3,000 pages of surgical materials, supporting various functions such as surgical phase identification and instrument detection.

CARES Copilot 1.0 is currently undergoing optimisation in seven hospitals in Beijing and can be integrated with medical devices. This development reflects China's efforts to catch up with global AI innovators by funding local initiatives, with CAIR's project being co-funded by the Hong Kong government's InnoHK research program. Last year, the Chinese government approved the rollout of the TaiChu model, another homegrown AI service by CAS, demonstrating the country's commitment to advancing AI technologies.

Professor Ming Feng from Peking Union Medical College Hospital highlights the potential of large language models (LLMs) like CARES in clinical settings, operating rooms, and research institutes. LLMs offer real-time support to medical staff, enhancing surgical safety by providing anatomical positioning information and enabling applications in surgery, image guidance, and robotics.