News

ELHS Institute Contributes to the NIH AIM-AHEAD Bridge2AI AI-READI Training Program
2025-02-05 00:00 Among the selected trainees, two distinguished participants—a university faculty member in computer science and a physician from Harvard Medical School—have chosen Dr. AJ Chen, Founder and Principal Investigator of ELHS Institute, as their mentor. We are honored to support this important NIH initiative with our expertise in fine-tuning open-source LLMs for clinical studies and applications.
ELHS Institute Publishes the First Comprehensive Review on the Democratization of Generative AI in Healthcare
2024-06-30 00:00 We are excited to announce the publication of our groundbreaking review on why and how generative AI (GenAI) is being democratized in healthcare. As ChatGPT-like AI continues to transform medicine, this review fills a critical gap by providing the first comprehensive analysis of the factors driving GenAI democratization, early evidence from the literature, and future directions for advancement.
ELHS Institute Publishes Novel ChatGPT Benchmarking Study in JAMIA
2023-12-20 00:00 In this study, we pioneered the use of live symptom-checking services, such as the Mayo Clinic Symptom Checker, to evaluate ChatGPT’s diagnostic accuracy. Testing across 194 diseases, we found that GPT-4 achieved nearly 80% accuracy, demonstrating its potential for clinical applications.
ELHS Institute Publishes Study on LHS Training with ChatGPT Co-Pilot in JAMA Network
2023-08-01 00:00 This study explores the capabilities and challenges of using ChatGPT as an educational assistant in LHS training. A key finding is that while GPT-4 provides mostly accurate scientific references, nearly all citations generated by GPT-3.5 were fictitious, highlighting the importance of verifying AI-generated information in academic and clinical settings.
ELHS Institute’s Pioneering Work on ML-LHS Simulation Published in Nature Scientific Reports
2022-10-26 00:00 To accelerate ML-LHS development, we pioneered the use of synthetic patient data to simulate ML-LHS units for disease risk prediction. This study demonstrated that ML-LHS can continuously improve as more real-world data accumulates over time.