I was wondering the other night whether it is possible to build a nerual net based only on the observable inputs and outputs + possibly incomplete (or complete)

I was wondering the other night whether it is possible to build a nerual net based only on the observable inputs and outputs + possibly incomplete (or complete) list of features of another neural net? This paper seems to say yes but I'm not sure