Title should read "Peak of Inflated Expectations Reached". I've worked on many health AI/ML projects. The last decade has produced tremendously powerful protot

Title should read "Peak of Inflated Expectations Reached". I've worked on many health AI/ML projects. The last decade has produced tremendously powerful prototypes which lay clear paths forward for productization. Sure, assay software that no one cares enough about hasn't been updated, but you better believe the medical apparatus as a whole will welcome any tool that increases throughput and increases margins. Automating radiology or facilitating radiology does just that. Sure radiologists might not like it, but radiologist do not operate health systems, MBAs do. For some perspective, medical devices take 3-7 years for approval. Most are not game-changing technologies. The ImageNet moment came in 2012. How can we reasonably expect to have functioning automated radiology only a decade from when we realized deep nets could classify cats and dogs?