Good points. But I genuinely wonder what role algorithm brittleness plays here.

“Fitting only to the test set” (see Andrew Ng quote in original article) is an acute concern in my circles: digital pathology in cancer research

See “Google’s medical AI was super accurate in a lab. Real life was a different story.”

https://www.technologyreview.com/2020/04/27/1000658/google-m...