I'm not saying every application is bad. It's just that most of the positive applications seem to mostly be data aggregation and cleaning problems, not hard ML problems. Most of the time it's really a simple algorithm with a lot of good data that you need, there's not much interesting work to do besides plugging things together