The Divine Comedy, Don Quixote, The Count of Monte Cristo, Madame Bovary, Crime and Punishment, Anna Karenina, Metamorphosis, One Hundred Years of Solitude.
These books represent a minute portion of the literature that has been translated into English and has deeply impacted both English literature and global literary culture.
Translation enables us to glimpse the shimmering fascination of variety, diversity and linguistic power, greatly broadening the range of literature available to us. It is also a fundamental way of sharing and retelling stories.
“ Translation is also another mode of authorship. A true literary translation is sensitive and stylish, a finespun work of mastery that often requires years of dedication and skill.”
Translation is also another mode of authorship. A true literary translation is sensitive and stylish, a finespun work of mastery that often requires years of dedication and skill. And it matters which translation you read, because different translators have their own style and often their own intentions for the text. There is no “perfect translation” of a text because each time it is transferred from one language to another, the translator must make compromises with vocabulary, meter, meaning and pace.
With the introduction of artificial intelligence (AI) models, and in particular large language models (LLMs), the landscape of translation is changing. Literary translation in particular has presented a complex and fascinating new ambition for AI software. It is a notoriously subtle and multi-layered endeavour — something seemingly requiring a certain human touch. How is AI being incorporated into the process?
Companies such as Nuanxed have begun to employ AI in the translation process. They use AI models to create drafts, which are then edited by human translators, reducing costs for publishers. While these tools are intended to make translation more affordable and efficient, many worry that translation work is being undercut and devalued.
AI-generated translations are often touted as time-savers. While some feel that they benefit from this accelerated process and save time by offloading more menial work, others assert that the time-consuming post-editing process is a significant drawback. Despite the relatively good quality of AI translations, the tendency to sometimes “hallucinate” information causes translators to spend more time correcting these issues and approach AI drafts with suspicion. Sean Bye, for example, a translator and executive director of the American Literary Translators Association, argues, “very often you’re better off redoing the whole thing from scratch” due to the copious amount of attention and effort required to alter these AI drafts.
AI translation services also have a problem with language representation. Because of the imbalance in training material available for AI models across languages, the quality of their translation outputs varies as well. Proponents of AI translation champion its potential to elevate minority voices, but there are also concerns about less nuanced translations being produced in less-represented languages. How is this representation actually going to be achieved without sufficient material with which to train the AI models? And how is the replacement of minority voices with AI models in any way progressive or uplifting?
Quality reduction is a serious concern when translating across all languages, even those that are well-represented in the AI training data. Liesl Yamaguchi, a French professor at U.C. Berkeley, claims, “A bad translation is worse than no translation because it’s going to block the way to a good translation being produced.” Once a mediocre translation is available, the push for another, better translation is often weak. With AI translations becoming ubiquitous, more ‘good enough’ translations of underrepresented languages will likely emerge, inhibiting human translators from creating more attentive, accurate translations.
Additionally, many translators raise the questions of ethics and purpose. Translator Julia Sanches, for example, stresses that AI translations “give the appearance that translation is instant, which devalues my labor, and also that it’s mediocre, which could make ‘good enough’ the new standard for the literary arts.”
The devaluation of the art of translation is what is really at stake. Automating the initial process often reduces professional translators to editors. While some point out that AI can be used as a tool rather than a replacement for human translators, many translators still raise concerns about machine translators taking over such a delicate operation and misrepresenting the time and dedication required to produce a literary translation. Post-editing is also lower-paid than translation, and translators who have been relegated to editing AI outputs often get paid less for an at times more challenging task.
Translation jobs are becoming more scarce because of competition with AI. A survey from the Society of Authors revealed that over a third of translators reported losing their jobs to AI in 2024, a number that has likely grown since. Similarly, around 4 in 10 translators reported that their income decreased due to the introduction of AI in the translation process. Translators are also concerned about the competition with AI for entry-level jobs, since many professional translators are being relegated to editing positions that still require a more advanced level of expertise.
So is human translation dying out? Potentially. But I’m not entirely convinced that it will ever be dead. We are still learning about how AI will impact the literary translation industry, and with the pace of AI’s advancement, it’s hard to predict what is to come. But the process of sharing stories across languages is so fundamental to the human condition that I find it hard to believe that it could ever be fully replaced.
The love that readers have for language is a crucial part of its dissemination; we delight in its “incredible nuances of usage,” says Yamaguchi, in “extraordinary translation.” So far, AI has not been able to create, or rather recreate, the nuanced linguistic depth of human translation: The human touch is still necessary for that connection.
When one day, it inevitably reaches a more sophisticated level of production, will human translators put away their pens and shelve their dictionaries? Translator Mark Polizzotti beautifully expresses his love for translation: “I do the work not so much to express as to discover what I think, where it will take me.”
Both translators and readers share this love of discovery, and a love of connection that we feel in the creation of a literary web: from the first origins of the story to its most recent permutation, from the author’s first spark of creativity to the latest reader of their creation. The process of translation is an art in itself that has inherent value beyond financial and temporal constraints. AI can facilitate the method of creating, but it can never drive the creation.
Ava Chambers PO ’28 is currently reading ‘The Decameron’ and ‘The Housemaid’. She enjoys watching movies, eating breakfast foods, and adding books to her ‘want to read’ list.
Facebook Comments