Medical artificial intelligence company Lunit has moved to expand the medical AI ecosystem by giving outside developers and medical staff a chance to use and test the medical-focused foundation model it developed in-house.
Lunit said on the 26th that it held the "Medical-Specialized Foundation Model Hackathon" on the 21st and 22nd at its headquarters in Gangnam, Seoul, with sponsorship from the Ministry of Science and ICT and the National IT Industry Promotion Agency (NIPA).
The event was organized to broaden the use of "L2-preview," a model Lunit developed through the "AI-Specialized Foundation Model" project, a national strategic initiative led by the science ministry, and to expand the medical AI ecosystem. About 70 people across 15 teams took part, including AI developers, university and graduate students, clinicians and engineers.
Participants used L2-preview to build medical question-and-answer systems and health-management chatbots and to carry out benchmark tasks. Clinicians and the foundation model's development team gave a comprehensive assessment of performance, medical accuracy and the potential for use in actual services. OpenAI provided ChatGPT Pro subscriptions and API usage credits to support the running of the competition and its awards.
The benchmark award went to Team AIM, which posted performance comparable to global frontier models on major medical-science benchmarks. The frontier award, which assesses potential for application in clinical settings, went to Team Nallyeohan Wegovy.
Lunit plans to connect the standout medical AI services identified at the hackathon to actual clinical settings. Using its network of partners, including hospitals, the company intends to offer clinical validation and technical support to the best entries and to carry these forward into hospital-based demonstrations and follow-up research and development.
"In the medical AI ecosystem, we need not just to release models but to create spaces where developers and clinicians can test them directly and build them together," said Yoo Dong-keun, Lunit's Chief AI Officer (CAIO). "We will provide support so that outstanding ideas and use cases can carry through to actual clinical settings."
