Google's generative artificial intelligence platform for healthcare is more than 90 percent accurate in making clinical decisions, on par with human providers, a July 12 study in Nature found.
Answers from the Med-PaLM AI tool aligned with the scientific consensus 92.6 percent of the time, compared to 92.9 percent for real-life clinicians, according to the study by Google researchers. The extent of the possible harm from the decisions lined up as well: 5.9 percent for Med-PaLM and 5.7 for clinicians. However, the large language model isn't quite ready to be deployed in clinical settings, the study's authors acknowledged.
"Although these results are promising, the medical domain is complex. Further evaluations are necessary, particularly along the dimensions of safety, equity and bias," they wrote. "Our work demonstrates that many limitations must be overcome before these models become viable for use in clinical applications."
Rochester, Minn.-based Mayo Clinic recently started testing Med-PaLM.