Yes. Just to be clear, I am here as an individual and not on behalf of Mayo, although that is where I work.
My own view of this, from being on the inside, is that there is a lot more claim-making and hopes than tangible and concrete results. Sometimes, whatever ends up working is much more following the steps of biostatistical rigour, which have been known or worked out over the past 50, 60 or 70 years, to get to an effective intervention that improves things in some ways. A ton of things that people are proposing may or may not fit what the actual health care needs are.
I would say that more biostatistics, and thinking of that as what ought to transform health care rather than labelling it as AI, is maybe a more helpful frame. There are works about this. I'd have to look them up. For example, a paper found that a lot of the AI tools for COVID were totally useless in the end. I think that's the case in a lot of studies that go back and look at it: Here's the AI that has been claimed to do something, and here's what actually happened. There is also a report from Data & Society talking about a successful implementation that was as much about the qualitative aspect and stakeholder engagement as it was about the actual model.
I would say that is where I am working and that is what I am working towards, but I would offer a lot of caution within that rhetoric.