Ha! I've used the exact same characterization.
Anyway, this seems entirely predictable given a couple different considerations.
1) People are most apt to effectively solve problems for which they have some context.
2) People who are underserved or otherwise have "real problems" don't often have much overlap with people who have discretionary money to pay for solutions.
So you end up with a selection bias in the pool of problem solvers (and thus the scope of addressable problems) and a selection bias in the expectation for investment returns. Which results in the best minds of our generation crunching all manner of machine learning models to make it so we don't have to do something mundane like selecting an outfit to wear that's color and style coordinated.