I’ll prepare to be impressed when it is shown by objective evidence and a double-blind study to actually be as reliable as billed. Until then, the cited article reads like promotional flackery for an unproven system being promoted by the very organization trying to sell it. I’ve seen way to many hype-fueled proclamations of chatbots ready to pump out legal briefs or solve long-unsolved problems mathematics fail to put lawyers and mathematicians out of work to take too much stock in what is essentially an extended press release.
The “promotion” is being done by The NEJM and Harvard … I am skeptical that they are trying to sell it. But correct double blind real world use trials are absent so far.
I’m not personally using AI professionally but it is being used broadly in our medical group and pushed hard by our leadership to facilitate documentation and billing efficiency (Abridge) and we are far from alone. I am resisting using it partly because smart phrases work well enough for me and writing the note for a more complex room is part of my thought process. Plus a noisy pediatric environment may be particularly challenging for ambient AI. I may try it out at some point though. Best case it works like having a very good scribe in the room.
The model is being developed by Harvard Medical School and five of the six authors of the study published in the NEJM are affiliated with Harvard Medical School (three of them with Mass Gen where the study was performed). This is not an independent study (understandably because it is still in development) which makes it far from definitive.