Maybe stated in a different way, how good do your priors have to be to learn and understand a language? With an infinite amount of training data, any language c

Maybe stated in a different way, how good do your priors have to be to learn and understand a language? With an infinite amount of training data, any language can be learned. The question is what prior assumptions are required to learn a language in a reasonable amount of time?