
Machine Translation Causes Man’s Arrest
At ITC, we have been vocal about the value and hopeful future of technologies that help bridge the gap between languages. It’s amazing how far technologies have come and how much value they offer in communication between cultures. We welcome the advances and look forward to seeing their increasing impact on the world. And, with that iterative progress,one should expect errors and setbacks.
Recently, Israel Police were notified of a concerning social media post due to an incorrect translation completed by Facebook’s translation AI. A man who posted “Good Morning”, with a picture of him standing by a bulldozer was arrested because Facebook mistranslated the post as “attack them” in Hebrew and “hurt them” in English. Without having the post first reviewed by an Arabic speaker,and the fact that such vehicles have been used in attacks in the past, police were concerned and arrested the man. After a few hours and some questioning, the man was released, and police understood that he meant no harm. The error seems to have originated from the use of a term for “good morning” that’s not an actual word in Arabic, but looks like the verb “to hurt”.
In this recent glitch of machine translation or AI used by Facebook, we can see that many technologies surrounding language services are far from accurate and struggle to consistently convey the appropriate context. Most of the time we see samples of this in humorous memes and awkward errors that cause little to no harm. However, this recent error does highlight the fact that “every word matters”. Context is important. And when it counts, qualified humans add value to ensure that the true message is accurately conveyed and understood. In this case, a quick review by a native Arabic speaker would have saved significant effort, time and embarrassment.
When talking about machine translations and related technology, we usually say the same thing over and over. It’s amazing technology, but it needs to be watched because it will surprise and embarrass you when you aren’t careful. It reminds me of an exchange from True Grit (2010). The young girl, Mattie Ross, tells Lucky Ned Peeper, “I would not be in this fixbut my gun misfired”. Ned responds sympathetically, “They will do it. It will embarrass you every time.”
If you are using machine translation for something important or of consequence, think again. For business, legal, professional, complex or even simple but lengthy content, we find that it can mangle the translation or interpretation beyond recognition and is completely unusable. Our proofreaders can spot a machine translation from a mile away because it just seems to be tied in knots.
That’s why we do what we do. Native only linguists, trained professionals and a process that reviews each translation with another native linguist. Additional measures are then taken to make sure everyone is doing their job and things are in order. That’s the only way we can stand behind translations in over 230 languages.