Imagine waking up one day to realize that you can communicate with anyone on the planet—regardless of where a person comes from, the language they speak, or how they talk.

Sound too good to be true? Well, that’s because it is—for now. But is a world without language barriers really so out of reach?

Photo by Rathish Gandhi, Unsplash

Language makes us human

Our ability to use language is what defines us as humans. How we communicate with each other and pass on complex and abstract ideas is remarkable.

With just a little vibration from my vocal chords, I can send soundwaves to your ear telling you what time we should meet for dinner, the latest political gossip, or how blockchain will change the world.

Something else that’s remarkable: there are 7,097 existing languages in the world.

But while the Internet and international travel have made the world feel smaller, language still remains a major blocker for global trade and cooperation.

In the European Union alone, the world’s second largest economy, we have 24 different languages. And this introduces some complications for the region’s economic growth.

For example, the Eurobarometer’s data shows that 19% of European Internet users never browse in a language other than their own. And only 15% of Europeans shop online from another EU Member State—compare this to three times as many (44%) who buy online in their own country.

This means that the EU is losing billions of euros every year because companies struggle to sell in new markets within the same region. According to Eurostat, an increase of just 5% in online sales in the region would generate an extra €195 billion every year for the EU economy.

That’s a lot of money.

So how do we solve a problem like translation?

Photo by Conor Luddy, Unsplash

The hard problem of translation

Translation isn’t easy—not for human beings, and especially not for machines.

This may surprise you, given headlines claiming that machine translation is “nearly indistinguishable” from professional quality.

But anyone who’s used Google Translate knows that it’s far from perfect—like when it detected “Ooga Booga Wooga” as Somali, or misled the Norwegian Olympic team into buying 15,000 eggs.

And it gets worse when you look at some other big brand examples.

Facebook had to apologize for getting a Palestinian construction worker in Jerusalem arrested for saying good morning. Why? Because their fully automatic translation service erroneously translated “يصبحهم” (or “yusbihuhum,” which means “good morning”) to “attack them.”

China’s WeChat app, a conversational interface with 900 million users, experienced a similar issue when it translated “black foreigner” to the N-word . In the end, they blamed their artificial intelligence software.

All of this happens because most machine translation engines are trained on simplistic and isolated sentences from Wikipedia and newswires. Because of this, they rarely capture the nuances and idiosyncrasies of human language. This might also be why a robot can’t be funny.

The future won’t be AI vs. Humans, but rather AI + Humans.

So while translation may just seem like decoding secret messages, it’s actually much more than that. There’s context, there’s culture, and there are things that only humans can understand—and these things are extremely difficult to pass on to machines.

Photo by Les Anderson, Unsplash.

Will machine translation ever be ready?

The latest advances in translation involve “Neural Machine Translation” (NMT), which has rapidly become the new state-of-the-art. It’s essentially a computer system which acts more like a brain by imitating biological neural networks. This lets it progressively learn and improve the more data it’s given.

But even NMT systems aren’t eliminating errors, let alone delivering a consistent tone of voice.

And that’s why we believe that a world without language barriers will only be possible if machines and humans come together.

The future won’t be AI vs. Humans, but rather AI + Humans.

Unbabel team

Humans are great at understanding context, grasping intentions, and asking for clarification to resolve ambiguities. Machines have quick access to word meanings in nearly every language.

Together they allow what we call “translation as a service”—fast, professional-level translation of video, customer support emails, and real-time chat.

In other words, we’re making language barriers a thing of the past for companies that want to go global.

And that can have a huge revenue impact in any business. In a world where personalization is key, it seems strange not to sell a product across multiple languages.

It’s like the old motto by former chancellor of the Federal Republic of Germany, Willy Brandt:

If I’m selling to you, I speak your language. But if I’m buying, dann müssen Sie Deutsch sprechen (then you have to speak German).

Breaking down language barriers may sound like an impossible mission, but we see plenty of reasons to believe otherwise. And we love a good challenge. Want to hear more from Vasco? Check out our interview with him on the future of human-machine conversations here.

Vasco Pedro is the cofounder and CEO of Unbabel, providing AI-powered, human-refined translations at scale. He’s also a surfer, traveler, Carnegie Mellon PhD, and father of four girls.

How you ask is everything.

Start creating