Honestly, the question is not, and has never been, "can machine learning replace radiologists/etc" (which won't happen in the foreseeable future). The question is, how can radiology/etc utilize ML to improve outcomes, decrease the cost of care, and broaden accessibility.