By: Nick Gambino
Google Translate has been around for nearly 20 years helping people navigate languages foreign to their own. Using numerous processes over the years to power translations, they’ve built the platform to support 133 languages.
Now, thanks to AI models, Google Translate is adding support for 110 new languages, bringing that total up to 243. That’s nearly double their capability in one fell swoop. The company used their very own PaLM 2 large AI language model to quickly bring Translate up to speed.
“From Cantonese to Q’eqchi’, these new languages represent more than 614 million speakers, opening up translations for around 8% of the world’s population,” the Senior Software Engineer at Google Translate, Isaac Caswell, explained in a Google blog post.
The newly added languages vary widely. Some of them are spoken by over 100 million people, while others are more niche or even rarely spoken natively but are part of “active revitalization efforts.”
Using the PaLM 2 AI language model helped tremendously, especially with languages that were closely connected. For example, languages like Marwadi and Awadhi have added support and those are related to the widely used Hindi.
Cantonese was finally added and apparently this was one of the most requested languages. The fact that it’s so closely related to Mandarin, and even has a lot of overlap when written, made it hard to develop AI models that could train Google Translate.
Punjabi, the number one language spoken in Pakistan, was also added. Around 25% of the new languages come from the African continent. They include languages like Wolof, Venda, Swati, Nko, Luo, Kikongo, Ga and Fon.
With 243 supported languages and more than 500 million people using the translation service daily, Google Translate has become as synonymous with crossing language barriers as their search engine has been with internet search. And now that they’re using advanced AI language models, I’m sure we’re going to see a ton more from the translation tool in the near future.
