Let's work in ##\Bbb R^2##, say. Suppose you had an equation ##F(x,y)=0##. Is there such a subset ## X\subseteq \Bbb R## such that for a fixed ##x\in X## there is a unique ##y## such that the equation is satisfied? If so, we say ##F## determines implicitly a mapping ##f:X\to\Bbb R## (satisfying ##F(x,f(x))=0, x\in X ##). We don't want to pick some funny weird subsets ##X##. We want it to be open so we could talk about differentiability (which is a very strong assumption - the course I took on optimal control theory Heavily relies on smoothness of the object function and the implicit function theorem becomes a powerful weapon - of course, it restricts the choice of ##F##, as well).