One more news from Meta. This time it is an AI model capable of translate 200 different languages – many of which are not supported by other commercial tools -. The model is part of an ambitious research and development project aimed at creating a “universal voice translator”. Automatic translation, in fact, allows Meta to better understand its users. But it could also be a useful technology for optimizing some future products, such as smart glasses.
Meta: The new AI model can translate over 200 different languages
Meta stands out once again for an incredibly ambitious project: an AI model capable of translating 200 different languages - although it seems that the quality of the translations of some of the lesser-known languages is inferior to the others -. More specifically, a all-encompassing model, capable of translating in more than 40,000 different directions. The company’s goal, in fact, is to include “low-resource languages” in the model, ie those with less than 1 million publicly available translated pairs of sentences. Among these many African and Indian languages usually not supported by commercial machine translation tools.
Angela Fan, Meta AI researcher, stated that the idea behind the Artificial Intelligence model is linked to the lack of attention often reserved for less widely used languages. “Translation doesn’t even work for the languages we speak, that’s why we started this project – said the researcher -. We have this motivation for inclusion like: ‘what would it take to produce translation technology that works for everyone’? “. And here is Meta’s answer to the question. According to Fan reports, Meta’s AI model is already in the testing phase to support a project that help the editors of Wikipedia to translate articles into other languages. And soon we will see the model integrated into the translation tools of the Meta platforms as well. Long last.