Yet another reason to be skeptical of reported performance in the medical AI space - maybe it's time to stop gaming the benchmarks and actually do some science. This is exactly why I've been harping on data quality for so long. https://www.reddit.com/user/ade17_in