The Limits of Self-Evidancing in Machine Learning

As a younger man I used to think that clearly there is and there's actually a mathematical bound as well as a sort of common sense physical bound in the set size of your brain. As you're maxing you're doing yourself evidencing it looks as if or implicitly you are generating very accurate predictions of your world because you're minimizing the complexity bit like Occam's razor so just by being self-evidancing  you can be too higher up really deep which means that you can have an overly expressive overparameterized network.