AI Language Translation
Microsoft researchers systematically evaluated GPT translations across 18 language pairs, concluding that the systems are pretty good on “high-resource” languages like English or Chinese, but still contain notable biases in the way they generate punctuation or favor smooth linguistic styles over technical accuracy. Hendy et al. (2023)
Jun 2024 Planet Money:
Despite great AI language translation “the number of people employed as interpreters and translators grew 11 percent between 2020 and 2023.” BLS expects that to continue to rise, and wages too. Reason: AI makes translation cheaper, which increases demand, including for humans checkers.
see (my tweet)
#China Jeff Ding points to this English translation of a Weixin post Did AI kill the translator? with an example of
In this translation, it seems correct to literally translate “limited evidence” as “evidence with some limits”, and the translation is also smooth. However, in an academic context, “limited evidence” does not simply mean that the amount of evidence is small, but that the amount and reliability of the evidence are so limited that it is not enough to support a certain conclusion.
and concludes that:
Excellent human translators, like chefs at Michelin three-star restaurants, will become ‘luxury service’ providers, serving only those customers who have very high requirements for translation quality.
Japanese Translation
Experienced Japanese expert Bob Myers May 2024 Translation in the Age of the Machine says
This is a technological revolution the likes of which no one in the language industry has seen in their entire careers. It dwarfs the impact of the last major innovations, translation memory and previous generations of MT. It would be extraordinarily surprising if the entire shape of the industry were not redrawn over the next five to ten years, with huge swaths of the tasks it currently performs being taken over by machines, in many cases operated not by them but by companies who had previously been their clients, squeezing them out completely. The only thing on the side of the lumbering language-industrial complex is pure inertia—clients too lazy or clueless to figure out how to redesign their business processes.
and he gives an example of how with the correct prompting, ChatGPT can correctly translate a literary novel.