Idea – Physical Cause An Effect As Deciding What Is Input And Output To A Machine Learning Network

Here is an interesting idea.

Could a machine learning network recognize the objects that form a machine learning network?

That is. Could it learn to recognize in detail the objects involved in cause and effect. From this could such a network generate a cause that has the target effect and so on. Maybe this could be used as an engineering tool.