In the context of globalization, machine translation plays an increasingly important role in bridging language barriers and promoting global communication. Machine translation technology is experiencing a paradigm shift from neural machine translation to large language models. Since the BERT pre-training model revolution, the GPT series models have continuously made breakthroughs in multilingual processing and translation, driving profound technological transformations.This research traces the historical turning point to the official launch of GPT models, and through academic papers, technical reports, and corporate news releases, systematically explores the driving mechanisms of machine translation technology during this shift. It examines the limitations of previous neural machine translation approaches and investigates the innovative mechanisms of large language models in translation practices, focusing on key technological breakthroughs in language models between 2018 and 2024. The research uniquely highlights that this shift differs from previous transitions due to the formation of an intermediate language. The embedding layer of large models converts source languages into unified word vectors comprehensible across all languages, which are then processed and transformed into target languages. The formation of this intermediate language and a certain degree of general intelligence gradually construct a "Babel" where translation between any languages becomes possible.The study aims to analyze the mechanism of technological shifts in machine translation, providing historical insights for the current development of machine translation technologies.

The paper focuses on the issue of establishing the IT field in the 1990s, the entry of strategic "players" into the scene in the Czech environment (mainly important companies, whether branches of multinational concerns or newly established domestic companies) and related unprecedented increase in the prestige of IT. One of the thematic axes of the presentation will be the gender issue – a certain feminization of the environment in the 1980s (the distancing of managers and male professionals from work with PCs that seemed to be burdensome) to the transformation of IT into a dynamic industry dominated by men-experts. The whole process will be reflected through interviews with people who actively worked and shaped the sphere of information technology in the 1990s, especially those who held managerial positions and thus became "carriers of change" in the Czech Republic.

This paper interrogates the early history of what we now call information theory, examining the antecedents to the post-war triumph of Claude Shannon's formalisation of electrical transmission as digital (discrete) forms of signalling through noise reduction. What kind of transmission 'theory' existed prior to this mid-century work? How did electrical engineers conceptualise the problem prior to the binary ascendancy of computing? I suggest that between the first decade of the twentieth century and the third, a qualitative shift in emphasis away from the materiality of the electric cable towards the nature of the signal arose in electrical engineering that concerned itself with telecommunications. Looking at telegraph transmission in the early 20th century in particular, I examine how engineers of the time conceived of and spoke about signal integrity, and how they understood the man-machine relations required to operate the telegraph equipment, and the struggles they encountered in attempting to automate our early telecommunications devices. The after-effects of this automation --- a redistribution of jobs away from expert transmission operators to arguably less-skilled administrative work ---- troubles the standard narrative of automation’s effect on labour.