If we could explain exactly how it works wouldn't we just write some "normal" logic to do what it does instead of creating a NN and having it learn on it's own.

If we could explain exactly how it works wouldn't we just write some "normal" logic to do what it does instead of creating a NN and having it learn on it's own. Even AI from 40 years ago is probably impossible to explain completely