a) One of the challenges in using autoencoders to model the dynamics of an environment is our inability to be certain that the important details are what’s being modeled