Translation after the machine: whose judgement, whose languages?
Two scenes from one field. In Stockholm, the Swedish Academy gives the Nobel Prize in Literature to Anne Carson and says her translations from classical Greek may be regarded as works in their own right. Meanwhile a European survey finds that more than a third of literary translators have considered leaving the profession, with AI a main reason, and the BBC asks how long humans will be needed at all.
Both scenes ask what translation is worth. Machine output keeps improving, yet one large study still saw more than a quarter of AI drafts corrected by people. The harder questions concern voice and inclusion: models trained mostly on English-only data leave out Africa's languages, bilingual speech is flagged as "broken", and AI adaptation can turn an egg hunt into an Eid activity.
Against this, communities write their own answers: Kafka in Réunion Creole, a 1,278-page Amazigh Quran translation, Indigenous languages on Apple keyboards. For Metaglossia, interculturation means that such questions are settled by communities and professionals together, not by tools alone.
📚 1. Literature, cultural exchange and the translator's craft
Translation shapes who is read, in which language and with what status.
🏅 1.1 Anne Carson wins the Nobel Prize in Literature
The Swedish Academy named the Canadian poet, essayist and translator, 76, and noted that her translations of Sappho, Sophocles and Euripides may be regarded as independent works in their own right.
🏆 1.2 Two Chinese-language books shortlisted for the US National Translation Awards
Shuang Xuetao's Hunter (translated by Jeremy Tiang) and Hon Lai Chu's Mending Bodies (Jacqueline Leung) are among 11 works from nine languages. Winners are announced on 23 October.
✊ 1.4 Binghamton hosts an IATIS workshop on translation and resistance
The first IATIS event in North America, on literary activism in conflict and solidarity, drew 21 speakers from eight countries. One keynote framed translation as memory and political labour.
A Mainichi column argues for keeping both vertical and horizontal writing, noting that over 30% of respondents to an official survey say they rarely meet vertical text in daily life.
The question is no longer whether machine translation works, but what it costs, whom it excludes and what humans still contribute.
⏳ 2.1 How long will humans be needed for translation?
In a Wikipedia translation study, 27% of AI-generated text was still corrected before publication, and 5 to 6% of errors changed the meaning. The BBC asks whether AI is now good enough to remove the human.
📉 2.2 37% of surveyed literary translators have considered a career change
A CEATL survey of 1,785 European translators reports that about 30% saw fewer or lower-paid commissions, and 67% of those offered post-editing refused it. The sample is voluntary, not representative.
🏫 2.3 ASU lab builds culturally sensitive multilingual AI
Standard AI can read Spanish–Quechua switching as faulty language. The lab pairs linguists and computer scientists to work from real, messy bilingual data.
An audit of three leading models adapting 60 maths problems into seven languages found diversity collapse, regional mix-ups and agreement on specific substitutions in only 33.5% of cases.
📱 2.5 Google Translate may add a "Smart Assist" mode
Spotted in the app's code, it would clean up filler words in spoken input, apparently in Conversation mode only. Timing and device availability are unclear.
🌐 3. Language rights, cultural identity and intercultural communication
Inclusion depends on data, standards and keyboards as much as goodwill, and on communities setting their own priorities.
🌍 3.1 How Africa shapes its own AI
A World Bank update says about 56% of open training data is English-only, leaving out some 2,000 African languages. It calls adaptation to local data, languages and problems the biggest payoff.
⌨️ 3.2 More Indigenous languages arrive on iPhone and iPad keyboards
Apple's September update added several Indigenous languages, including Blackfoot. Schools use the keyboards to teach endangered languages such as Cherokee.
📜 3.3 Morocco publishes the Quran's meanings in Amazigh (Chleuh)
A 1,278-page edition from the Mohammed VI Foundation, made by scholars of the Souss region through weekly collective review and three proofreadings, is presented as the first official translation of its kind.
🇪🇪 3.4 Estonia adopts its first plain language standard
In force since early October, it puts the reader first with broad principles rather than strict rules, and is meant for anyone who writes for work, not only language professionals.
A translator-intern at a Chinese AI start-up argues that the hardest thing to translate between China and its Eurasian partners is pace, and that a polite "let's think about it" can mean no. An opinion piece issued as a press release.
🎯 4. Professional development and the dynamics of the field
Tools, prizes, research and training shape what the profession can do next.
⛪ 4.1 Breeze Translate: live captions and spoken translation for churches
Any phone can show captions in about 200 languages and play audio in about 75. It is used on six continents, which its lead credits to recent advances in AI.
The award for translators from Ukrainian accepts applications from 12 October to 4 December. The winner receives 3,000 euros (after tax) at a ceremony in spring 2027.
Linguist John Hawkins explains how ambiguity, and the process of disambiguation, decides high-stakes court cases where he has served as expert witness.
A review of a South African study of horses and welfare NGOs extends translation theory beyond human language, drawing on a biosemiotic view of translation.
Human judgement is being re-priced, not replaced. Machines draft quickly and still need correction, while many translators report fewer or lower-paid commissions.
Inclusion is built into data and standards. English-dominated training data, plain language rules and Indigenous keyboards all decide who is understood.
Translation is creative and political. From a Nobel citation to Creole neologisms, the work is more than transfer.
Interculturation goes beyond equivalence. Pace, indirectness and cultural defaults shape meaning as much as vocabulary.
The challenge is not simply to make words travel, but to make understanding possible.
💬 Our call to the global language community
We invite translators, interpreters, researchers, educators, students, publishers, language-technology specialists and language-rights advocates to join the conversation.
Who should decide when a machine translation is good enough?
How can translators and interpreters secure fair recognition and pay as AI spreads?
How can every language community shape the data and tools that represent it?
How can translation help us move towards genuine mutual understanding?
Every language carries a world, and every genuine act of translation creates an opportunity for worlds to meet.