Deep Learning - The Biggest Problem in Machine Learning

The idea of using historical data, and relying on that seems to go against the very thing i was taught to say when sold mutual funds as a way to support my family. The thing ive had to always mention was, past performance is no guarantee of future success. And so now there are two mistakes built in from the very beginning. One is thinking you can learn about now from studying before. And also just looking at correlations as though their causation as that's a double barrelled problem y there.And the third problem appears to me it's a terrible problem when you don't know how the machine has learned what it claims to have learned,. You rely on its its answers. But