While the press regularly uses artificial intelligence to translate foreign articles – as evidenced by an article published in “Libération” on September 13, on the “flowery language” of Kamala Harris – translator Bérengère Viennot, author of “La langue de Trump” (Les Arènes) warns that the practice is not without danger. Translation is a profession, its effects are not without consequences on society and the time and money saved by AI raises the question of whether we should sacrifice the quality of articles for their quantity.
On September 13, the daily Release has published the translation from an article of the Washington Post, “under the supervision of our journalists, then edited by the editorial staff”. This article addresses the fact that US Vice President Kamala Harris, a candidate in the November 5 presidential election against Donald Trump, has a more than flowery language. I will cut the suspense short immediately: despite the reported human intervention, it is a completely rotten translation.
A more than questionable translation
I’ll spare you an exhaustive explanation of the text, a glance at the first paragraph will suffice. It creaks. At the Democratic convention, Vice President Kamala Harris spoke about advice her mother gave her : “Never do anything half-assed” [qu’on pourrait traduire par “ne torche jamais rien” pour garder un peu la vulgarité, mais on se contentera de “ne fais jamais les choses à moitié”, plus proche du sens initial, N.D.L.R.]. And if there’s one thing Harris doesn’t do half-heartedly, it’s his use of swear words. »
Rather than translating the expression “half-assed”, Liberation leaves it in English and explains it. Probably because artificial intelligence came up with something bland – Deepl, the internet’s go-to translation software, offers the ” never do things by halves ” who according to Liberation would be “ closer to the original meaning », understand: this is what said the machine, we are not translators (indeed), so that must mean that but we feel that something is missing.
ALSO READ: “Let’s not be afraid to speak French”: when the Academy castigates anglicisms, which are prevalent among the elites
We can imagine that since Trump, journalists have been scalded by the translation of the slip-ups of American politicians, and for good reason. In the early days of his mandate, they often took the full brunt of a language that sometimes disconcerted even translators because it violated the laws of semantics, and they often came up with distressing translations.
A small example: the “ You’re in such good shape ” Trump threw at Brigitte Macron the first time he met her, when he failed to hide his surprise at a woman over sixty who still fit his criteria of beauty. Most journalists had translated it as “ You are in good physical shape ” which was actually a slob phrase closer to ” What shapes! ” and whose subtext was rather ” You are awfully well preserved for an old lady. “.
Cheap sub-translation
Because to translate a human thought, you have to understand it – but artificial intelligence only translates words, not ideas. It doesn’t know that Trump is sexist, a bimbo lover, and violent in his relationship with women. The translator knows this, and translates accordingly – not the words, but what’s behind them, based on the original style.
To return to the half-assed by Kamala Harris, ” never burn anything » is not entirely bad – just off the mark. While torcher does indeed mean “to botch” a job, on the other hand, it is not idiomatic, that is to say that in French, we simply would not say that. On the other hand, in English, the expression raises a smile. It is a little vulgar, but its humorous side makes the pill go down. We could more happily opt (for example) for “ never bust your ass halfway when you do something “And there is no need to justify yourself – a press article is not a written assignment.
ALSO READ: “Allowing the French language to say everything”: how the State wants to slow the rise of Franglais
Finally, I could expand on the following sentence: “ And if there’s one thing Harris doesn’t do half-heartedly, it’s his use of swear words. ” but I will simply point out that ” make half use of swear words ” is closer to a drunken late-night babble than to correct French. The rest of the paper is similar and the various attempts to translate the ” fuck ” of the original version are not a success – one could even say that they did not work. The entire text is peppered with sub-translations, half-translations or non-translations, and completely misses the style and rhythm of the original.
I can hear you from here. Ah, those Luddites who cry out against robots because they are afraid of losing their jobs! Well, yes. The round of complaints from neuron workers is partly due to the fact that we are afraid of starving to death. Of course, we have to move with the times and this will not be the first time in human history that a cohort of workers will be replaced by machines. However, until now, the revolutions that have seen robots replace humans have allowed them to free themselves collectively. They tended to relieve human societies of the constraints that oblige individuals to waste their lives earning a living to satisfy the vital needs of the body: eating, clothing, sleeping under cover.
Saving time at the expense of quality
The revolution we are going through today is very intellectual. The Internet promises to put all the knowledge in the world at the service of all and with automatic translation and writing, journalistic or literary, the prospects are endless. To have all the books in the world translated, for nothing, in record time, and to offer them to the whole world: what a fabulous prospect! To have novels and articles written by machines: what a time saver!
The problem is that for now, this perspective is only quantitative. As we saw with the lamentable translation produced by Releaseif the result is readable in the literal sense of the term, it betrays the original thought. It dulls it, softens it, makes it insipid. Even “ revised by the editorial staff “(!) this text would not deserve an average, or barely (I rate it high), in a class of writing skills applied to translation.
Because to translate a human thought, you have to understand it – but artificial intelligence only translates words, not ideas.
And I won’t even mention the cases where the machine translation is downright wrong. This happens more often than you might think – but to know this, you need to have the skills of a translator… To settle for “post-editing”, that is to say to entrust someone whose job it is not to check that the machine-translated text is correct, is to expose yourself to delivering to the reader a writing that is at best totally bland, at worst that downright betrays the original.
What do you want to feed yourself with?
So beyond the editor-in-chief’s approach, who says that it seems to be talking about the same thing for cheap, the time has come for you, the reader, to ask yourself a question. What do you want? YOU to feed ?
Will you choose quantity, that is to say an abundance of articles, books, translations, most often free and cheap to produce (let us emphasize that the text proposed by Liberation is not yet freely accessible) intellectually mediocre, even fallacious? Or that of quality, even if it means sacrificing overabundance and, often, free access, for the benefit of depth of reflection and quality of language? Basically, do you prefer to feed your intelligence with junk food all you can eat or good food at fixed times?
ALSO READ: French language: when was the first anglicism?
The trend that we, translators, see coming with texts produced by artificial intelligence is that of a downward leveling of language and, therefore, of reflection and intelligence. Is it useful to recall that language is the indispensable tool for intellectual freedom, to reread 1984 to remember that the limits of language are those of thought?
So yes, we are afraid of losing our jobs, and that is not fun. But we know how not to fall into the trap of the standardization of thinking favored by artificial intelligence. As professionals of the text, we know how to protect ourselves from textually transmissible diseases. And if we want to protect the quality of writing, it is also to protect you, the readers, from the advent of an intellectual universe where artificial intelligence, nourished by its own mediocrity, will starve your thinking.
