Not so big tobacco Deal to roll Philip Morris and Altria into $200bn giant sparks analyst doubt

Regardless of language barriers, AI is making instant communication accessible to all, find James Cook and Hasan Chowdhury

Afrikaans into Khmer? No problem. Hungarian into Punjabi? Easy peasy. Once the realm of science fiction, technology that allows people to communicate instantly with each other using different languages is now a reality – and it’s poised to change the world.

That at least is the bold claim made by experts amid rapid advances in translation technology – based on sophisticated artificial intelligence and machine learning.

The $50bn (£40bn) global market for translation technology is booming and it promises to upend everything from education and travel to the world of business, work and diplomacy.

Students can dump the dictionaries, travellers can communicate with locals naturally and politicians no longer have to rely on a human interpreter.

In theory, at least, they should allow for more effortless communication – and the spread of new ideas across different cultures around the world.

Eventually, it could become so advanced that people won’t even know translation is taking place, says Dr Saihong Li, professor of translation studies at the University of Stirling. “Translation technology really plays a crucial role in politics, travel and in everyday life.”

As the accuracy of real-time translation improves, so the cost of this technology has fallen sharply. Google, for instance, offers £160 Pixel Buds headphones that allows users to translate languages instantaneously.

Similar technology has been developed by Waverly Labs to let two people who speak different languages understand each other using a £200 earpiece. Meanwhile, Amazon and Microsoft both have their own translation engines, which can be used

to convert text to a foreign language. Apps like itranslate and Translate Voice also allow anyone to have a powerful AI translation tool in their pocket, often for little to no cost.

But what does this all mean for the future of languages? And what are the negative implications?

Li claims we may no longer need to learn languages out of necessity. “I don’t think machines will completely replace humans but nobody can learn all languages and for those languages we can use technology. It’s a pity that fewer people are learning languages … second language learning for human beings [will be] self-satisfaction.”

While fewer people may learn languages, the need for translation services is only set to increase.

According to a recent report from the University of New South Wales, English speakers still dominate the internet – but only just. They comprise 28pc of internet users, followed by Chinese speakers who make up 23pc and Spanish speakers at 8pc. However, when it comes to the online content available to these users, English leads at 56pc, with a plunge to Russian and German (both 6pc), Japanese and Spanish (5pc), and Chinese at 3pc.

The discrepancy between these figures shows just how much translation tools are becoming an essential for the digital age. Improving the quality of machine translations is big business for technology companies.

Clare Suttie, the director of translation company Atlas Translations, uses her company’s mix of computerised machine translation and humans to translate text.

Suttie’s company translated the online listings for Boot Bananas, a shoe insert that helps to stop shoes smelling, into several European languages, turning the business from a living room project into a British company exporting products around the world.

“The people who do really well on Amazon will spend a small amount of money, it might be £100, to get a quality translation of their shopfront and they’ll notice that they get a return on that investment because their sales go up,” Suttie says.

Raisa Mcnab, chief of the Association of Translation Companies, says many UK companies exporting their products overseas are using automated services to translate product information. But she sees “massive volumes of data,

‘I don’t think machines will completely replace humans’

never before translated, that are benefiting from machine translation”. She says: “It’s the content companies would like to have translated but where there isn’t and will never be a budget for human translation.”

It isn’t just businesses that are benefiting. The Google Translate app’s camera function means travellers can point their smartphones at signposts in foreign languages and know what they say. Meanwhile students in schools could gain access to education that would otherwise be foreign to them and politicians could exercise their diplomatic flair with a host of nations.

But for all its promising applications, translation technology still has some major issues. In 2018, for instance, Facebook was forced to apologise after it mistakenly translated Indonesian comments where users said they hope people survive a fatal earthquake as people saying “congratulations”, and then displayed colourful animated balloons and confetti around the comments. A major leap in translation technology occurred several years ago when businesses such as Google started using neural networks in their translation systems.

Neural networks mimic the brains of humans and animals in order to teach themselves how to do tasks, whether that’s playing video games or translating an Amazon listing from English to Italian.

“It’s like taking a kid and teaching them until they become an expert brain surgeon,” says Ofer Shoshan, chief executive of One Hour Translation.

His business is arranging a conference in China next month on the subject of neural networks in translations, and he says he was surprised at the “massive” demand for tickets from Chinese tech businesses developing their own translation systems. There’s also a high level of demand for translations from English into Chinese, say experts, as firms look to do business in China and work with manufacturers in the country.

But it’s Asian languages, including Chinese, Japanese and Korean, where many machine translation services are the weakest. “There’s still a lot of work to be done in these languages in order

‘There’s still a lot of work to be done in [some] languages to get good results’

to get good results. The structure is different,” says Shoshan.

Even though artificial intelligence systems may struggle with some languages, Shoshan firmly believes that mass layoffs of human translators are coming in several years.

Is Suttie worried about a possible collapse of her industry? “Not at all, no,” she says. “Anything but.” However, she adds: “There’s a large number of translation companies. Possibly some of them might disappear.”

Whatever business advantages may be presented by AI, the simple pleasure of learning languages may mean it remains a perennial activity for people.

For Colin Watkins, of Duolingo, the uptake of languages on the app-based learning service has been a testament to the idea that people will still seek ways to learn languages because of the “visceral” experience of being able to communicate with people directly.

“AI is not in that position where it can understand nuance and tone and even regional peculiarities,” he says. “How do you measure sentiment in language?”