But boring problems -are- hard problems. I find very little is 'hard' in the sense of being difficult but ultimately overcomeable. Most things are either impossible (but can be fun to think about), or rather simple but requiring a lot of time (learning a new thing), or easy and quick (which we do and don't think about), or easy but long, and thus boring. And it's these latter that are, to me, 'hardest', because they're the ones I, and basically every one of my colleagues, least want to do. But as you say, they also can be some of the most valuable.
You mention machine learning and brilliant algorithms? Did you clean your data? Did you refine the parameters to the algorithm? Did you try different algorithms and compare? Machine learning is -boring-. Like, the things it enables are exciting, once it's working it's exciting, but the process is boring as hell. And I find that aspect every bit as 'hard' as understanding the underlying math.
