What is the future of translations in the AI era?

With the advent of AI and its penetration into practically all layers of modern society, many voices have been against its use, as it would impact job vacancies.

According to a recent analysis, the IMF found that almost 40% of global employment is exposed to AI, with slightly higher percentages in the advanced economies. The automation brought about by AI could entail lower labour demand, lower wages, reduced employment and, in some cases, jobs may disappear altogether.

This means that emerging markets are less exposed to AI without the proper infrastructure or skilled personnel that may hone and harness the benefits of AI.

How AI impacts the translation market

AI-powered translation and localization tools are far from being perfect but specialists work on improving their efficiency. The attempt to transpose natural languages into a world of strict rules that govern the software behind the AI has been successful to a certain extent. The newest NMTs (Neural Machine Translation) have become much better, and this is largely due to the exposure of the NMT to huge datasets of human translations.

On the other hand, automation software (such as CAT Tools, MTs) and AI designed specifically for translation purposes have positively contributed to the translators’ activity, by reducing their workload and improving their overall delivery capacity.

Despite the loss of jobs to automation, new jobs emerge, in which translators and localization experts will be required to train and manage these AI-powered tools and ensure that AI outputs provide better quality. Some experts believe that translators and localization experts will be forced to specialize to remain competitive.

Caveats of using AI-powered tools in translation

Although I have used AI-powered tools extensively over the past years, I couldn’t help but notice some hidden dangers.

Because the AI does the translation for us (which is far from perfect, even with the best tools available),  the translators must shift the focus from the actual translation to proofreading. This may prove less of a challenge for expert translators but for the beginners, things might be entirely different. Inexperienced translators will tend to “adopt” the AI’s syntactical and vocabulary habits that are not quite correct and, as such, will end up misusing the language. Therefore, their translation output may not enjoy 100% accuracy, which, in many cases, leads to material or intellectual losses as a result thereof.

Another major issue of AI training is that it uses (many times) incorrect source language datasets – in terms of syntax, vocabulary and structure. Since AI algorithms are based on mathematical rules, they will eventually output flawed translations that must be proofread by human translators.

An incident with the US immigration department that happened in 2019 shed light on the limits of AI-powered translation tools.

Carlos, who is Afro-Indigenous, speaks Portuguese but does not read or write it. Staff at the Calexico, California, detention center spoke only English or Spanish. The staff used an artificial intelligence-powered voice-translation tool to interpret what Carlos was saying, but the system didn’t pick up or understand his regional accent or dialect. So Carlos spent six months in Ice detention unable to meaningfully communicate with anyone.

In that time, he had no clear idea of why he was being detained or where his family was. When he sought medical care for his high blood pressure and for Covid, the nurses had trouble understanding him, he said. Spanish-speaking fellow detainees helped to fill out his asylum application, but the translation tool they used failed to produce an accurate account. It didn’t recognize Belo Horizonte as the name of one of the cities Carlos had lived in, instead translating it literally to “beautiful horizon”. And in response to a question about the mistreatment he suffered, the application read: “YES THE GANGUE DO BURACAO TO SHOOT DEAD MY SON, IN THE POLICE I WAS SLAPPED.” (The Guardian, https://www.theguardian.com/us-news/2023/sep/07/asylum-seekers-ai-translation-apps)

Although the DHS has contracts established with several translation companies, Border Protection officers were instructed to use Google Translate to vet refugee applications.

As tempting as it may be to use AI-powered translation tools to cut costs, the outcome may become troublesome if they are used in an unsupervised manner (in our case, the discrepancies voided the entire immigration file.

I have said this numerous times: don’t use AI translation tools unless you’re effectively proficient in the languages you translate in/from. Otherwise, the lack of experience combined with the apparent sense of security conveyed by the AI tool may lead to unwanted consequences.

Even though AI-powered translation tools help us speed entire processes in an attempt to provide faster results, this shouldn’t be done at the expense of accuracy and quality, because, in the long run, the data used to feed the AI will become less and less dependable and the results will only get worse.