Yet another AI reproducibility / reliability / generalizability crisis, this time in medicine

https://www.nature.com/articles/d41586-023-00023-2

The reproducibility issues that haunt health-care AI

Health-care systems are rolling out artificial-intelligence tools for diagnosis and monitoring. But how reliable are the models?

@garymarcus naively makes sense to me, humans and the conditions we suffer from are super high dimensional (sepsis is a great example of this) -- too easy to overfit without massive, clean datasets. As mentioned in the article, those datasets don't exist in this field due to privacy & regulatory issues creating data silos.