Lost in AI Translation: Why the Human Touch Still Matters

With literary output surging and the pool of efficient and reliable professional translators shrinking, can AI bridge the gap quickly enough? While machine translation lacks human cultural nuance and local context, is ‘good enough’ becoming the new standard?

While AI significantly reduces translation costs and labor, concerns are growing over its long-term risks. Disincentivizing professional human translators could backfire, ultimately compromising quality and reducing overall translation output.

This problem is especially prevalent in Europe, where Eesti Raamat Editor-in-Chief Piret Lemetti told Eesti Rahvusringhääling (ERR), the Estonian Public Broadcasting, "If you cannot publish the book at the same time it appears in English, then you have already missed the train." Readers need to feel a connection between their native translation and the alternative language.

Translators want to ensure all countries are able to get the same amount of materials in real time but face pressure of meeting the demanding workload. They say that most of the classic translations have been maintained, but now translators are saying they can't keep up with newer literature translations, as according to an update by Zippia, 4 million books were published in one year, with 2.3 million of them being self-published titles, and this number is only growing.

And so are translation rights fees and author fees, which force many more publishers to turn to AI to help shoulder some of the weight and volume of the book industry.

It was announced last year that Taylor & Francis would be using AI to help them translate more academic articles for a quicker turnaround time and had their program go through a year-long training. 

Mark Robinson, Taylor & Francis corporate media relations manager, said to Publishers Weekly, "We are interested in publishing translations of high-quality books across a broad range of advanced learning topics and formats, including academic monographs, textbooks and resources for practitioners." 

This comes after the public scrutiny the publisher faced in 2024 for signing a deal with Microsoft AI for 10 million to sell access to its authors' research. As well, it was reported just last month that Amazon was using AI to translate literary classics and then destroying them. The move underscored translators' fears that AI is being treated as a replacement for the craft rather than a tool for it, and that the books being lost in the process are the classics they've spent decades preserving.

Yet, the confidence to rely on technology to understand cultural connections and contexts while translating for hundreds of other books and languages isn't quite strong, especially for those trying to build up public trust again. 

The works of new foreign authors also have a small audience of approximately making them unprioritized amongst translators and pushed to the wayside as other new mainstream books come out that are catching popularity quickly. Varrak Editor-in-Chief Krista Kaer recalls that Nordic literature is becoming increasingly popular and needs to be published, and at the same time so is Korean literature, but there aren’t enough translators to go around, and so a tough decision has to be made.

Lemetti echoes that, telling ERR that the pressure on translators amounts to a "devaluation of the translator's work. Essentially, the translator is reduced to a glorified editor. I do not understand what the point of this is," and "If readers cannot associate authors with someone or something, they are not very inclined to buy the book," she said.

These new technological changes overall underscore the translators’ fears that AI is being treated as a replacement for the craft rather than a tool for it, and that the books being lost in the process are the classics they've spent decades preserving.