Animated evolution of recurrent neural network state space (clevercode.blogspot.com)

So let's see. Parity is kind of a large version of XOR right? Does a multilayer NN fail at this task or does the topology have to expand with the input dimension for that?