GPT-4 outperformed 99.98% of simulated human readers in diagnosing complex clinical cases

GPT-4 outperformed 99.98% of simulated humans, diagnosing 52.7% of complex cases in a Denmark study using 38 clinical case challenges. Compared diagnostic results of GPT-4 and 248,614 online medical journal readers. Study included diverse diseases, ages, and genders. Research indicates GPT-4's clinical promise, but emphasizes need for further trials and considerations for global benefit, ethical implications, and regulations.

Opens on MobiHealthNews