When to assume neural networks can solve a problem

The question: “What are the problems we should assume can be solved with machine learning?”, or even narrower and more focused on current developments “What are the problems we should assume a neural network is able to solve?”, is one I haven’t seen addressed much.