With the advancement of recording technology in various parts of biology, we can now observe many elements in a biological system at the same time.  For instance we can measure the expression level of many genes using gene microarrays or spike trains from many neurons using multi-electrode arrays.  Given this, an important and interesting question to ask is "can we use such data to infer interactions between elements of the system?".  In this talk, I will describe toy versions of this problem.  I will briefly describe how it can be done for an equilibrium Ising model, that is the inverse Ising problem; then I discuss inferring the interactions of a non-equilibrium Sherrington-Kirkpatrick model showing how Dynamical Mean-Field theory (naive mean field, TAP and exact MF) can be developed and exploited for inferring network connectivity.