On the dynamics of multi agent nonlinear filtering and learning