If you ask academic machine learning experts about the things that annoy them, high up the list is going to be overblown headlines about how machines are beating humans at some task where that is completely untrue. This is partially because reality is already so damn amazing there is no need for hyperbole. Most of Atari is solved. Professional transcriptionists lose to voice recongition systems. Object recognition has been counted on the machine side of the tally for years (albeit with a few more reservations). Considering the headlines we see, this may surprise many people. For someone who watches the medical AI space, it seems like a day can't go by without some new article reporting on a new piece of research in which the journalists say machines are outperforming human doctors. I'm sure anyone who stumbles on this blog has seen many of them. I didn't even have to search for these. Almost all of them are still at the top of my Twitter feed.

MRI scans of patients with radiation necrosis (above) and cancer recurrence (below) are shown in the left column. Close-ups in the center column show the regions are indistinguishable on routine scans. Radiomic descriptors unearth subtle differences showing radiation necrosis, in the upper right panel, has less heterogeneity, shown in blue, compared to cancer recurrence, in the lower right, which has a much higher degree of heterogeneity, shown in red.