National Taiwan University Hospital to go multimodal in major AI development

National Taiwan University Hospital is is moving to the next phase of its major AI development to focus on multiple data types.

The hospital recently purchased two new supercomputers from NVIDIA, which will help accelerate the development of smart healthcare applications.

This includes projects that develop multimodal large language models for optimizing operational processes and improving the quality of patient care, NTUH said in a statement.

THE BIG TREND

NTUH first acquired NVIDIA’s AI supercomputers in 2020, leading to the development of several generative AI and extensive reality-based applications.

For example, the Center of Intelligent Healthcare has developed LLMs for ICD-10 auto-coding, auto-generating health check reports, auto-generating transcriptions of telemedicine consultations, reporting emergency voice recordings, and extracting key points from pathology reports. NTUH’s IT office has also applied LLMs to summarize medical records, generate reports, collect unstructured data and answer medical questions. These LLMs are integrated into NTUH’s health information system.

With its supercomputing capabilities, the hospital has also developed a VR platform for surgical training called OpVerse.

Meanwhile, two leading electronics manufacturers in Taiwan recently announced projects to ambitiously build what could be the country’s largest supercomputing capacity. Foxconn, in collaboration with NVIDIA, aims to develop the fastest AI supercomputer in Taiwan with more than 90 exaflops of expected AI performance, potentially supporting cancer research and LLM development. Last month, Foxlink also unveiled its first supercomputing facility, Ubilink – also powered by NVIDIA – which claims to have Taiwan’s highest computing power yet.

Outside of Taiwan, two major healthcare clusters in Singapore, SingHealth and the National University Health System, have used supercomputers in recent years to accelerate their development of AI applications, including one genAI-based chatbot and digital twin technology for monitoring disease outbreaks.