Yeah, that's right. This class of problems (what are adequate priors, how to learn them automatically) is under the field of representation learning in ML.