This is why I tell everyone who is using AI for any purpose that the first thing to do is ask it questions you know the answer to very well. Because that’s how you find out how reliable the answers are. When you see how confidently it says things that are *almost* right, but not
@erinbanks.bsky.socialMy current employer recently got rid of UpToDate access and instead is providing OpenEvidence that integrates into our EHR. Sounds cool, yeah? Except that OE routinely provides clinical summaries with details that are categorically untrue. I’ve witnessed this multiple times.