生成式人工智能时代对译者任务的再思考

打开文本图片集

中图分类号：H319.5 文献标识码：A 文章编号：1000-0100（2026）01-0007-8

DOI编码：10.16263/j. cnki.23-1071/h.2026.01.004

Rethinking the Translator's Task in the Era of Generative AI

Li De-jun

（Collge of International Studies，National University of Defense Technology，Nanjing 21O039，China）

With the advancementof large language models（LLMs），theircross-lingual translation capabilities havecontinuously improved，triggering widespreaddiscusionabouttheroleandfutureofhumantranslators.Thispaperexaminesthestrengthsand limitationsof generativeAI（GAI）fromtheperspectiveofitsoperationalmechanismsandexploresthepotential spaceforhuman translatorsinlightoftheessenceoftraslation.ThecurrentresearchindicatesthathileA-poweredtranslationsystemsbasedon LLMsholdsignificantadvantagesinknowledgeretention，effciency，andmultilingualcoverage，heystillfallshortinhandling low-resource texts，lackawarenessof broaderlinguisticandsocioculturalcontexts，producerepetitiveand homogenizedoutputs duetodatauniformity，and exibit“centralized”translation tendencies stemmingfromrandomized weight-baseddecision-making.Evenaspe-trainingdataincreasesandsomeof thecurrent problemsofintellgenttranslationsystems may eventualyberesolved，GAIcotfullyplacemanstoughout tentiretaslationpross.Intelligenttrasatioat tasedtheistoricityoftranslation，anditsoutputscanonlyberegardedasintermediateproducts（Res intermedia），astheyfocusononlysemantic elementswhilelacking ethicaland political dimensions.Onlyafterthetranslatorreflectson theseintermediateproducts withaneyetowardsthe futurecanthetranslationbecomeafinalproduct（Resfinalis）.Duetothe“contradictory”natureof the politicsandethicsoftranslation，thereisnoconsistentandunchangingstandardforevaluatingthequalityoftranslations.There processing of intermediate translationproductsandtheselectionofoutputresultsarecucialtothetraslationprocess.Therefore， humanoersightremains idispensable incomplexcontexts，specialized tasks，and creative endeavors soas toupholdthe prnciples of“fidelity in meaning”“recapturing of linguistic beauty” and“pursuit of appropriateness”.

Key words：generative artificial intellgence；large languagemodels；Al-powered translationsystems；translator'stask；es sence of translation；centralization of translation

1引言

2022年岁末，OpenAI公司发布了ChatGPT，人工智能从此进入一个全新的阶段，或称为“生成式人工智能”（GAI）阶段。（剩余15256字）