People have been trying to incorporate speech recognition into healthcare for a long time. Ages ago I worked at a company which was trying to do it for consultants writing letters to patients' GPs. And the idea was that this could reduce the need to employ thousands of secretaries and save money in the healthcare system. Writing up patient notes is not the same thing but also takes hours of people's time which they could be spending doing something else important.

So this isn't a need that's been pulled out of thin air. It also shouldn't be one that's difficult to assess the efficacy of: we should have a pretty good idea of how to measure how effective speech transcription is, and we should be able to assess summaries for completeness and errors.

But medical data needs to be protected, and any process that relies on storing and processing it with US big tech is suspect there. Which it may well already be, but this is another opportunity for it...