TRANSLATOR COMPETENCIES IN THE ERA OF GENERATIVE AI

The article examines the transformation of professional translator competencies in the era of widespread adoption of generative neural networks. The relevance of the study stems from the fact that neural machine translation (NMT) systems and large language models (LLMs) produce increasingly fluent but potentially unreliable texts, rendering the traditional skill of post-editing insufficient for ensuring translation quality. The research is based on an analysis of contemporary scholarly works on the problem of hallucinations in machine translation, as well as on empirical material — translations produced by various NMT systems and LLMs, which were analyzed by the authors to identify stable types of errors. Based on the analysis, an extended typology of errors and hallucinations characteristic of generative models is proposed (detached, oscillatory, factual, detail, style hallucination, among others). The necessity of a shift in translator training from narrowly focused post-editing instruction to the development of comprehensive competencies for editing translations using Generative AI is substantiated. These competencies include textological analysis, critical thinking, prompt engineering, and data management skills. An updated model of professional training is proposed, aimed at shaping a new type of translator — a meaning interpreter and quality auditor capable of effectively interacting with generative technologies.