x
Progress update: synthetic models of natural data — LessWrong