This chapter offers a reflection on translation within the linguistic frame but beyond the human and explores the logics of neural machine translation services, particularly those of Google and Amazon (AWS) Translate. Research in Natural Language Processing is quickly evolving but the promise of NMT is that it is both scalable and localizable; more to the point, this mode promises, in Amazon’s terms, to ‘easily translate massive volumes of user-generated content in real-time.’ The chapter investigates how NMT is particularly important for the academic discourse on translation with respect to the phenomenon of ‘zero-shot translation,’ which is translation between languages with which the system has no prior experience. What are the implications of translational practice without a pre-existing framework, or in Barthes’ language, a ‘pre-ordained state of language,’ for the field of translation studies? What new understandings of language might be prompted, even necessitated, by zero-shot translation?