Good morning dear translation buyers (and everyone else for that matter)! Are you ready for the next
AI/MT translation fail? It’s an hilarious one – so buckle up!

As always: please take 2 minutes to read this post below, and then imagine you are the client in this situation, with your website, newsletter, your device’s user manual, in a regulatory setting, in the subtitles of your expensive marketing video, in your eLearning course, your app interface, your support centre, etc.

This time it’s DeepL Pro – yes, it’s AI. Here’s what DeepL says about it: “With DeepL, you’ll never have to compromise on quality again. Powered by neural networks and the latest AI innovations, our technology captures even the slightest nuances and reproduces them in translation, unlike any other service.”

Ok, then. I was recently working on the translation of an article from German into French for a medical journal (it was their specific wish I use DeepL Pro, which is integrated into my TMS). The article discussed the current landscape of paediatric care provision in a given country. The German term in the article title is “Kinderärztliche Versorgung” – roughly translated in English as “paediatric coverage.”

You know what DeepL Pro spat out as a translation for “Kinderärztliche Versorgung”? Wait for it…
“The food supply in paediatricians”. 😮🙄😐🤭😆😂

“Versorgung” is a complex word to translate from German, because depending on the context, it can have a whole range of meanings – supply, provision, procurement, feeding, coverage, services… But clearly the “technology (that) captures even the slightest nuances and reproduces them in translation” is unable to comprehend this common term – and it’s indeed so nuanced that it suggests cannibalising poor
paediatricians.

Imagine if the client, a medical journal, had not hired a professional and had just plugged the German article into DeepL Pro, copy-pasted the French output, and off it goes: journal published, printed and distributed to hundreds of medical professionals – bearing this title. Just imagine the faces of their audience of doctors and clinicians (not to mention their paediatrician readers, running for their lives)

Imagine now that you are the Director of this medical journal, and this article is posted on your website or printed in your product flyer. Hard to tell just how massively it would damage your credibility, right?

And yet, many companies keep trusting DeepL Pro, ChatGPT, Claude, etc. blindly for their translation needs – meaning they simply input their texts into one of these apps, and then use the raw output without bothering to have it checked by a language professional or at least check it themselves or internally – playing with fire and taking reputational and credibility risks.

We LPs and translators are the professionals. We know what these tools are capable of as well as their limitations, and we know how to work with them, while ensuring you a proper result that will not hurt your credibility (or any innocent paediatricians).

So don’t trust AI, don’t trust MT – trust your human translation service provider.

(No paediatricians were harmed to generate this image – which OF course was AI-generated, just for the irony of it ;))

Discover our smart solutions for the medical and healthcare sector >>

Image credits: generated using DALL-E from OpenAI, just for the beauty of the irony (ok, and also because good luck finding a stock image of a physician INSIDE a plate!)