This article proposes a comprehensive view of the development of machine translation methods and the circumstances surrounding the emergence of the advanced translation systems and technologies: from decryption, through rule-based machine translation, to statistical methods of translation and, finally, neural machine translation that provides a “decryption” of the text on a new and deeper level. The author concludes that after seventy years of discussions and controversy the research consensus has shifted towards the original and seemingly long-abandoned scientific programme.