This article aims to redefine the role of corpus as that of an intersection between human translation (HT) and machine translation (MT). Contemporary translation studies recognize MT's rapid technological progress, but also its many shortfalls, particularly on the front that it is not translation-actual but mimicry. In this article, we first examine machine translation's history, then the creative differences between human translation, generative AI, and MT. Machine learning for generative AI is based on a corpus, that is a collection of text written by humans. This corpus can also serve as the repository of creative writing for MT to learn from, by means beyond the traditional concept of archival language text collection and machine-imitation. In this way, corpus serves as the intersection between HT and MT, and also as the means to improve the creative capacity of MT.