Translations - Doomed to AI or Sleeping Giant?

Oftentimes, like Galen Erso in Rogue One, we have found ourselves burning the midnight oil and encountering an Imperial Shuttle landing in our office. Then, out steps a bad-tempered admiral, demanding us to come build the AI Death Star. We refuse. Why? We’re busy with a translations rollout - the AI Death Star can wait!

Now, we’ve done a lot of translations in all sorts of languages—dozens of translators around the world, hundreds of thousands of words, myriad voice-over sessions, and countless creative adaptations. But whenever we’re busy with these, I think to myself - are we partaking in a dying art form here? Are we the radio DJs, the travel agents, the clerks behind the desk at a Blockbuster? I generally end my musing by accepting that translation is a dying art with value as a niche offering to specific clients.?

But the numbers disagree.

In 2021, the translations market size was $61 billion and is growing at a CAGR of 5%, which is great for a mature market. On the other hand, the market for machine translation in 2021 was $800 million and is growing at a CAGR of 30%. Wait, AI translation tools aren’t worth even a fraction of traditional translation? So, traditional translation is alive and well! Not so fast -?

Disruptor AI technology often has a disappointing actual value when you take a hard look at the numbers. But this is not because the tech is bad. It’s because the tech is way newer and less established than we can comprehend - think how long translators have been doing their thing versus how long AI translation tools have been viable. And, importantly, the AI tools are significantly more affordable than the traditional approaches they disrupt. So the bet is that in a few years, we’ll see these numbers swing dramatically. Well, the reality is more interesting.

While translations seem simple, doing them well is one of the most complicated streams of work you can take on. Especially when dealing with large quantities, needing to create efficiencies, and working in a specific brand tone of voice. You end up with a multi-stage creation and approval workflow. The reality is that AI is viable across multiple, but not all, stages of the translations workflow. So you won’t see AI taking over translations. But you will see companies like Stratia implementing it effectively in their workflow.

And there is the plot twist. As the admiral walked away, he had no idea the real reason we refused to build the AI Death Star was that we’d stolen the technology on our way out and were implementing it in our little AI-powered translations farm.