A study conducted by a trio of pediatricians at Cohen Children’s Medical Center in New York found that the diagnostic skills of ChatGPT are lacking. They tested ChatGPT by asking it to diagnose 100 random pediatric case studies, but found that it was only correct 17 times, and even then some of those were still wrong. The researchers noted that ChatGPT is not ready to be used as a diagnostic tool, but could potentially be useful for administrative tasks or other purposes. They suggest that more selective training could improve its diagnostic abilities. The study was reported in the journal JAMA Pediatrics.
Source link
ChatGPT found to have very low success rate in diagnosing pediatric case studies
