Back in 2010, in the wake of Haiti’s earthquake, Nicole “Mac” McClelland, an American writer who doesn’t speak French, met a handsome French UN peacekeeper called Nico Ansel. 

A year and a half later, and after one accidental break-up, when Nico mistakenly told Nicole after an early date that he didn’t want to see her again, they got married, thanks in no small part to help from Google Translate. 

Armed only with a smartphone and an internet connection, Google Translate means that travellers, businesspeople and new arrivals in a foreign country can make connections with other people, wherever they come from and wherever they are. 

Last October, 10-year old Amanda Moore passed a Google-Translated note to a new classmate, Rafael Anaya. 

Rafael had just moved to California from Michoacán in Mexico and didn’t speak English. 

The folded note read “Would you like to sit with me today?” Her mother posted on Facebook about how the two had connected despite their lack of a mutual language – and the post went viral. And Amanda and Rafael are now friends.  

Google Translate and ice cream 

Sometimes these are short relationships. 

In Rio for the 2016 Olympics, gold medal-winner Usain Bolt was fresh from celebrating his 30^th birthday, and used Google Translate to connect with local student Jady Duarte. That was a short-lived connection: Bolt’s girlfriend back in London was less than impressed when photos of their encounter hit the internet.

Google Translate can help start off significant relationships, too. 

Bea Longworth, 36, works for tech firm NVIDIA, and frequently travels from her home in Oxfordshire in the UK to Munich.  

That’s where her new German-speaking stepson lives – and Google Translate is helping them build their relationship despite the language barrier. 

His Canadian grandparents, meanwhile, combine tech with more old-fashioned approaches.  

Together they navigated Munich's Tierpark — a 99-acre zoo — “with only the aid of Google and emergency ice cream when things got tricky,” says Bea.

The Hebrew for spinach 

Steven, who’s a student in Tel Aviv, decided to cook dinner one evening for his new girlfriend. 

His lessons in Hebrew had not equipped him for identifying on supermarket shelves the ingredients in non-Roman characters for spinach and ricotta lasagne, he says. 

After wandering the aisles looking for spinach, he hit on using Google Translate. (If you're ever in a similar position, look for תרד).  

“We then had a similar experience with walnuts,” he says. 

The effort paid off: the couple is now living together, and, says Steven: “I'm getting better at Hebrew - especially food.” 

And Google Translate relationships sometimes lead to polyglot Google Translate weddings. 

Andrew Johnston, a Scottish-Australian finance consultant, married his Russian wife Evgenia Kulagina – choosing Italy as the venue.  

“Evgenia was fine, but the rest of us were all on Google Translate for at least one language,” laughs wedding guest Laura Dance.

Lost in Google Translation

Google Translate is great when it works, but it can lead to some big cross-cultural misunderstandings – and they’re not always Google’s fault.

Steven Levitt, the author of Freakonomics, was travelling in Germany in 2014.

Levitt, a University of Chicago economics professor, wanted to be gracious to someone his host.

Having typed in what he thought was a compliment for his host, Hedwig, he was a bit puzzled by what Google Translate returned.

“What Google Translate gave back to me seemed a bit odd, but then again, German often seems odd to me,” he says.

He was even more nonplussed when Hedwig replied that what he’d said was so far from real German that it would be better if he never said it again.

It turned out that he had in fact asked Google to translate “I had such a mice time with you.”

And recounts Jonathan Jewell, a 41-year old educator in London, “a Spanish girl I liked never talked to me again after I translated something I wanted to say to her through it”.

Though he admits: “I'll never know if it was lost in translation, or if the fidelity of the translation was the issue.”

Caught in a neural net 

Internet translation has spread beyond Google. It’s now a built-in feature in WhatsApp, which Beyers Coedzee, who is South African, relies on to communicate with his girlfriend Vera Lamnci, a Russian speaker from Nizhny Novgorod, even when they are side by side.  

Meanwhile, Microsoft Translator has added languages like Mexico's Querétaro Otomí, Laos's Hmong Daw, and (useful if you need it) Klingon. 

But convenient, free translation in your pocket began when Google Translate was launched in April 2006. 

Initially, Google's translation software used multilingual transcripts from the United Nations and the European Parliament.  

Comparing the same texts in different languages, the tool would gather statistical data about how words and phrases corresponded in different languages – an approach called statistical machine translation. 

It works best with languages that are close linguistically, like Spanish and Portuguese, and ones where it has a rich source of data, such as English. 

Frequent users noticed a sudden improvement in Google's translations in November 2016.  

The reason was the roll-out of a new technology called neural machine translation. 

Neural networks aren't programmed for particular tasks, but instead learn as they go without supervision. 

Rather than being set up with linguistic rules, they use algorithms to generate their own rules.  

They also take on entire sentences at a time, whereas statistical translation uses words and phrases like building blocks. 

Neural translation can get up to speed with a brand new language in a week, and compared with phrase-based translation, reduces errors by 60 per cent, researchers at Google said in a paper.

Though statistical machine translation has been around a while, and goes back to efforts in the late 1980s and early 1990s at IBM's Thomas J Watson research centre, neural machine translation has really come of age just in the last year. 

And it is catching on. 

Facebook announced in a blog post recently that it is moving its translation “from using phrase-based machine translation models to neural networks”. 

And one New York start-up, Waverly Labs, has crowd-funded a translating device, which will use neural translation.

The device, which has already sold $5m-worth in pre-orders, shipped its Pilot device in December. It comes with two earpieces, one for each person in a conversation, with the earpieces linking to a smartphone via Bluetooth. 

“It's a process in three steps - from speech recognition to machine translation to speech synthesis, so each has its challenges,” says Marion Guerriero, Waverly Labs’ vice-president.

Communicating with people with whom you don't share a common language using technology has long been a science fiction staple. 

Star Trek features a Universal Translator, which unlike virtually every other thing on a spaceship, never breaks down.

 Arthur Dent in The Hitchhiker's Guide to the Galaxy learns to understand other species by inserting in his ear a small yellow Babel Fish, which the Guide calls “probably the oddest thing in the Universe”. 

In Doctor Who, it is the TARDIS's translation circuit that does the trick. 

But now instant translation is finally moving out of science fiction and into real life – and it’s set to break down more and more language barriers. 

No more “Hovercraft full of eels”, then. 

Meanwhile, we don't yet know if “mice to see you” has quite caught on in Germany.