{"id":657205,"date":"2026-05-08T07:07:09","date_gmt":"2026-05-08T07:07:09","guid":{"rendered":"https:\/\/www.newsbeep.com\/au\/657205\/"},"modified":"2026-05-08T07:07:09","modified_gmt":"2026-05-08T07:07:09","slug":"being-human-helps-despite-rise-of-ai-is-there-still-hope-for-europes-translators-ai-artificial-intelligence","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/au\/657205\/","title":{"rendered":"\u2018Being human helps\u2019: despite rise of AI is there still hope for Europe\u2019s translators? | AI (artificial intelligence)"},"content":{"rendered":"

In February 2022, while he was plugging away at rendering the US writer Dana Spiotta\u2019s novel Wayward<\/a> into French, the literary translator Yoann Gentric decided he needed a bit of light relief. He would test whether AI could put him out of work.<\/p>\n

Gentric had been grappling with a short non-verbal sentence that described the book\u2019s protagonist\u2019s feelings upon opening a window: \u201cBright, sharp night air, bracing.\u201d He put the prompt into DeepL, a neural-network-powered machine translation engine that regularly<\/a> outperforms Google Translate in accuracy assessments.<\/p>\n

The proposed translation was reassuring, with his job security in mind: L\u2019air de la nuit, vif et vif, \u00e9tait vivifiant (The night air, lively and lively, was enlivening.) AI had translated the sentence\u2019s meaning but was seemingly unaware that the repetitions rendered the line absurd. It was far inferior to his own translation that would be published in the book a year later: L\u2019air pur et piquant de la nuit, vivifiant.<\/p>\n

Yoann Gentric tested AI translations in 2022 and 2026 and found very different results.<\/p>\n

When Gentric repeated his experiment this spring, however, the outcome made him feel less at ease: L\u2019air nocturne \u00e9tait vif, pur et vivifiant, DeepL suggested this time. The online translator still lost the sentence\u2019s stylistic trait by adding a verb, but it had learned to use three different words that even had a musical ring to them. \u201cI don\u2019t know if it\u2019s just chance or a fine-tuned algorithm at work, but nocturne and pur is not bad,\u201d said Gentric.<\/p>\n

Chatbots running on large language models (LLMs) \u2013 neural networks trained on vast amounts of text to generate natural-sounding language \u2013 are rapidly infiltrating every aspect of our work and leisure lives. But few professional sectors are being disrupted by the technology as rapidly as the translation industry in Europe<\/a>, home to more than 200 languages and a booming tech sector.<\/p>\n

According to a recent joint survey<\/a> by the French authors\u2019 societies ADAGP and the Soci\u00e9t\u00e9 des Gens de Lettres, 79% of translators believe the rise of AI \u201cposes a threat of replacing all or part of their work\u201d. In Britain, a 2025 survey<\/a> found that 84% of translators questioned expected lower demand for human translation, resulting in lower pay.<\/p>\n

Those fears concern the future, but for many translators the nature of their work has already changed. Laura Radosh, a Berlin-based German-to-English translator, used to get about four job requests per month from clients including universities, professors and museums. Last year, the number of offers dropped to one each month.<\/p>\n

Many of them were \u201cpost-editing\u201d jobs, which required her to correct texts that had already been run through a machine-translation engine. \u201cPost-editing took me as much time as translating from scratch,\u201d said Radosh.<\/p>\n

Far less creatively fulfilling than translating from scratch, post-editing is also less well-paid: usually compensated by the hour rather than by the page or by the book, it is paid \u201cat unacceptable rates considering the work involved\u201d, according to the French translators\u2019 association. In Germany, publishers have been found to offer typical rates of two to eight euros per page \u2013 a quarter of the average pay for translating a page from scratch.<\/p>\n

But rates for regular technical translations have tumbled too. \u201cI got offered a job at 60 cent[s] a line,\u201d said Radosh. \u201cBefore then, 80 cent[s] was the lowest rate I had ever come across.\u201d<\/p>\n

Even before the advent of LLMs, translation was a precarious profession: a recent survey by the German translators association Vd\u00dc found that the average income for literary translators \u2013 traditionally at the lower-paid end of the sector \u2013 was as little as \u20ac20,363 euros per annum before tax<\/a>. But the latest changes in the industry mean that for many translators, the numbers no longer stack up \u2013 Radosh recently took a part-time job doing book-keeping for an NGO.<\/p>\n

Marco Trombetti, the co-founder and CEO of the machine translation company Translated, said: \u201cWithout help, the human brain basically is able to produce about 3,000 words a day of translation. Beginners will manage about 1,500, the best translator in the world may manage 6,000, but the variation is not that big.\u201d<\/p>\n

The cost of human translation, he argued, had until now been defined by the number of neurons we have in the brain. \u201cThat\u2019s around 100bn,\u201d Trombetti said. \u201cBut if we change that, then we change the unit economics of translation.\u201d<\/p>\n

Yet the speed of technological change is also revealing what human translators still do best. For one, many machine translators still struggle with context. The German-British academic publisher Springer Nature offers its authors the option to have their books auto-translated into other languages for free, but in spite of assurances of subsequent \u201chuman checks\u201d, this process has in the past led to comical results.<\/p>\n

In 2024, Springer Nature machine-translated into German an English-language book by a group of Indian academics called \u2018Capital\u2019 in the East: Reflections on Marx<\/a>. In the chapter headings<\/a>, however, the machine translator DeepL had rendered \u201ccapital\u201d not as Kapital in the intended sense, but Hauptstadt, meaning \u201ccapital city\u201d.<\/p>\n

A spokesperson for Springer Nature said in a statement: \u201cOur AI\u2011supported translation is human\u2011led and reviewed by professional editors. Errors like this are rare and regrettable, and this instance relates to a limited pilot that has since ended.\u201d<\/p>\n

J\u00f6rn Cambreleng, the director of Atlas, a French organisation promoting literary translation, said: \u201cMachine translation is not creative. These systems are built to produce sentences that are generic, sentences that have been said before or sound like they have been said before. Whereas good human translators strive to put into words something that has never been said before.\u201d<\/p>\n

Katy Derbyshire: \u2018I understand what someone might scream when they hit their toe on the bed frame \u2013 an algorithm doesn\u2019t.\u2019 Photograph: Nane Diehl<\/p>\n

One of the ironies of the upheaval is that literary translation now appears to be a comparatively safer career choice than its technical counterpart.<\/p>\n

The HarperCollins-owned imprint Harlequin France has confirmed that it is working with a French communications agency, Fluent Planet, to produce translations that are generated by AI software and then post-edited by humans, although for now such trial runs are confined to the pulpier reaches of the market: Harlequin\u2019s titles include A Mistress\u2019s Confession and The Embrace of a Prince.<\/p>\n

In Germany, where the total number of new published books has been gradually declining year on year, literature in translation has held up remarkably well, with 8,765 books in translation published in 2024 making up a historically high 15% of the overall output<\/a>. Increasingly, authors are also contractually obliging their publishers not to use AI in the translation process, said Marieke Heimburger, a Danish-to-German translator who chairs Vd\u00dc.<\/p>\n

\u201cAI really cannot do dialogue,\u201d said Katy Derbyshire, a Berlin-based translator who has rendered into English novels by Clemens Meyer, Christa Wolf and others. \u201cWhen you are translating from scratch, you learn to understand the characters and their motivations, and you\u2019re constantly adjusting them in your head \u2013 to individual situations, but also to genre. The dialogue that AI came up with just didn\u2019t suit the character description at all.\u201d<\/p>\n

Being human helped the translation process, she added. \u201cMy body has experienced all the pain and the joy that literature strives to convey. I understand what someone might scream when they hit their toe on the bed frame \u2013 an algorithm doesn\u2019t.\u201d<\/p>\n

Fernando Prieto Ramos, of the University of Geneva\u2019s faculty of translation and interpreting, said his centre had noticed a drop in applications to translation courses three years ago, when the rise of generative AI fuelled the hype around machine translation. \u201cBut the trend is gradually reverting again with a more diversified training offer,\u201d he said.<\/p>\n

Even people who develop machine translation software concede there are tasks that remain beyond their product\u2019s reach. \u201cIf in Italian I say Solo tre parole: non sei solo, then a literal translation into English would be \u2018Just three words: you are not alone,\u2019\u201d said Trombetti, who founded Translated in 1999. \u201cBut you\u2019ve ended up with four words, not three. That\u2019s something that machine translation still struggles with.\u201d<\/p>\n

Heimburger said: \u201cI am not really scared of AI, because I know it cannot do what I can do. What I am afraid of is the people who think that AI can do my job.\u201d<\/p>\n","protected":false},"excerpt":{"rendered":"In February 2022, while he was plugging away at rendering the US writer Dana Spiotta\u2019s novel Wayward into…\n","protected":false},"author":2,"featured_media":657206,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[256,254,255,64,63,105],"class_list":["post-657205","post","type-post","status-publish","format-standard","has-post-thumbnail","category-artificial-intelligence","tag-ai","tag-artificial-intelligence","tag-artificialintelligence","tag-au","tag-australia","tag-technology"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/posts\/657205","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/comments?post=657205"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/posts\/657205\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/media\/657206"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/media?parent=657205"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/categories?post=657205"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/tags?post=657205"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}