Artificial intelligence in translation: advantages and limitations

Advances in artificial intelligence (AI) development have significantly improved the quality of machine translation, resulting in more accurate and correct translated texts. However, the ability of AI to replace human translators completely is under consideration. The aim of the study is to trace the degree of adequacy of AI-assisted translation and identify its limitations. The authors explored the text of Vladimir Putin's interview to the American journalist Tucker Carlson given on 9 February 2024 and its English translations: the official translation posted on the Kremlin website and the translation obtained using the DeepL platform, which is currently regarded as the most accurate AI translator. The research focused on the semantic, grammatical, stylistic and pragmatic adequacy of the two translated versions. The comparative analysis revealed a number of limitations of neural machine translation, indicating the insufficient ability of AI to make the right choice between the meanings of polysemantic words, recognize figurative meanings of words, adequately interpret phraseological expressions and find their translation correspondences, discern the syntactic structure of the sentences, note expressive and other stylistic nuances, take into account the linguistic and sociocultural context. The results of the study prove the need for post-machine editing. They may find application in neurolinguistics and contribute to the development of neural machine translation.