It was on my trip to Japan in 2023 that I realized how powerful Google Translate (really any translator) is. I did have some previous knowledge of Japanese, but I wasn’t fluent enough to speak with the native Japanese. I then realized that I could use Google Translate. This blog post will discuss the GNMT (Google Neural Machine Translation). It will also briefly discuss the benefits and uses of translating applications and websites.

Google Translate in modern days has revealed itself as a very integral part of our everyday lives. People from all over the world such as tourists, foreigners, and others of that ilk can quite flawlessly communicate with each other with the assistance of Google Translate. A primary example is a tourist (like in my case). When I was in Japan, I needed to use Google Translate to inform the native speakers of things like the fact that I didn’t eat meat or ask how to pay for different things. Ultimately, we can tell (from everyday use), that Google Translate is a crucial and useful tool that helps us communicate with others.

The GNMT is a Machine Translation method that is much more effective than traditional translation methods (the use of ‘lingua francas’ that serve as language “bridges”). It can directly translate languages such as Japanese and Korean, while using English, not as a bridge, but as a common language. The algorithm goes something like this: English is translated with Korean and English is translated with Japanese, then the two English translations are matched to find the correct match between Korean and Japanese.

Throughout the process of simplifying translation, GNMT was able to create its language. The GNMT system started taking shortcuts to further simplify translation methods. This has formed a sort of ‘interlingua’ which is a language that combines multiple languages to communicate. The ‘interlingua’ was created when the system tried to use different languages and combine them to find ways to translate directly from language to language (or even through one singular language, the ‘interlingua’). Researchers have yet to decode this ‘interlingua’, but they can understand the fundamentals as shown above.

This is very intriguing to me as I never thought that AI and the Google Translate system would ever be able to create something as complex as its language. This has broadened my view on Google Translate and the different languages out there. It has overall shown me that to create bridges on top of bridges there eventually has to become one dominant language (which I always thought would be English). It turns out that the GNMT system has created a counter-argument to my original thinking and that it has shown that it could be possible for all languages to combine some time in the future.

Sources

NationalAdmin. “Google’s AI Translation Tool Seems to Have Invented Its Own Secret Internal Language – National.” National, December 17, 2019. https://www.national.edu/2017/03/24/googles-ai-translation-tool-seems-to-have-invented-its-own-secret-internal-language.

Image Credit: Getty Images

West, NurPhoto, Bloomberg, NurPhoto, Anadolu, Sopa Images, Fotograzia, et al. “Google Translate – Application Close up on Apple iPhone XR Stock Photo,” n.d. https://www.gettyimages.com/photos/google-translate.