12 March 2024
A spectre is haunting Europe
Ein Gespenst geht um in Europa – K… (A spectre is haunting Europe…)
Only this time, the K stands for KI, rather than for “Kommunismus”
Translators have been worried for a while that they might be overtaken and replaced by machines. And they know best of course. The general public, some of whom had perhaps tried out Google Translate when it first appeared on the scene were more sceptical (“It’s rubbish”, “Not usable”, “It does not understand humour”, “It cannot cope with idiomatic language” …). But now, it seems, everyone who cares about language must be worried.
An article published in the Tagesanzeiger (TA), a Swiss daily founded in 1893 and the newspaper with the largest number of subscribers in Switzerland, carried a pertinent article on 11 March 2024. Under the provocative title “Foreign-language learning? No longer needed”, it discusses the plight of people currently working as translators. It quotes one computer linguist lamenting the fact that we are “threatened with an enormous cultural loss”.
The TA journalists put DeepL and Chat-GPT to the test. They choose two passages, one journalistic one, a review of a play, and a literary one, a passage from Julie Zeh’s bestseller Unter Leuten. The first included two idiomatic expressions “aufs Tapet bringen” and “durch den Kakao ziehen”, plus terminology belonging to the gender debate. The source was German in both cases. The theatre review was translated into Finnish, then into Indonesian, then into Polish, Rumanian, Swahili, and then back into German. You can look at all these translations here:
Their conclusion regarding the final back-translation into German: amazingly good! The content of the review had been kept intact with no loss of information; it was perfectly intelligible, the machine had cracked the idioms and found perfectly good solutions. Some nuances, the TA team concedes, were lost, but nothing serious. No more serious than what previous similar experiment involving humans had yielded.
Julie Zeh’s text was translated into more “conventional languages”, i.e. languages where there is much more material available for the engine to learn from. It was first translated into English, then French, Spanish and finally back into German. Here, they say, the result was even better, nuances were preserved, as was the speech melody and the rhythm.
Clearly, the TA team were surprised (aghast?) by these results. If, they say, the results are this good right now, what is going to happen in the near future? Will the profession of the translator indeed become redundant?
They discussed the “spectre of AI” with a number of literary translators in Switzerland and Germany. The reactions from them indicated that they did not feel that KI was an acute danger (yet) and that many publishing houses were very quality-conscious, some of them making it a condition that their literary translators work without a digital helper. At the same time they did admit that some of their colleagues used AI as a kind of online dictionary, or perhaps to get a proposal about a particular passage. This brings up the question of whether in such cases they are under an obligation to inform their publisher of this extra help… One of the translators who was interviewed is convinced that large publishers still place more value on their professional image and on high quality rather than free-of-charge, super-fast translations.
What these translators fear is that the image of translation as an art and a profession will be impaired and that some publishers will start using AI as a means of reducing the fees they pay. This fear, and the threat of being downgraded to post-editors has prompted the literary translators in Germany, Austria and Switzerland to publish the “Manifest for Human Language”.
Its preamble goes like this:
Literary translators are already experiencing how the automation of intellectual work and human language affects their field and society as a whole: both art and democracy are under threat. We, the literary translators’ associations of Germany, Austria and Switzerland, along with our supporters, warn against these developments and urgently request remedial measures.
In our view, text-generating artificial intelligence is a technology that poses a systemic risk, and strong regulation is absolutely imperative.
It then lists a set of demands, followed by answers to the questions 1) What is Literary Translation, 2) What are Language Bots?” and 3) How is the Sustainability of Our Profession Threatened – and What Are the Broader Implications?” This is followed by the actual Manifest, which comprises 6 points.
If you have read this blog, I would urge you to go the website, read the manifest, and, if you feel strongly about the topic, submit your vote.
Click button bellow to apply to work as a freelancer.
Apply as a freelancer
