>10 years on and i still haven't seen a single question answered with clustering

Zuhause 😁

a Stulle

John «most of my published research is wrong» Ioannidis

Couldn't agree with this more. They also don't have particularly good editorial oversight, as @stephenkb.bsky.social has pointed out for BBC news.

Thanks to all of the fantastic co-authors, none of whom are on here AFAIK :)

The full paper is here: www.gstatic.com/amie/multimo...

More interesting: AMIE was also rated better on a whole host of criteria, by both the patient actors, and by specialists. Patient actors had a better experience, felt that they AMIE was better at addressing their concerns about things shown in images, and interpreted images correctly more often.

When comparing diagnostic accuracy, AMIE performed better:

We compared multimodal AMIE to clinicians in an OSCE study: Trained patient actors held (blinded) conversations with AMIE and PCPs based on cases in which image information was crucial.
We used three image types: dermatology pics, scans and snaps of lab reports, and printouts of ECGs.

Multimodal AMIE is built on a state-aware reasoning framework in which the AI navigates a diagnostic conversation by asking at every state: What's known about the case? What is the current differential? And what information could help narrow down that differential?

Gone in 120 seconds