The HealthBench test can't possibly tell us the critical factor: How humans would respond to chatbots under real-world conditions.



Source: ADnet
Source Link: https://www.zdnet.com/article/openais-healthbench-shows-ais-medical-advice-is-improving-but-who-will-listen/