Mark Zuckerberg, CEO of Meta, announced a major development of the “No Language Left Behind” project initiative on his official Facebook page. Facebook. The main objective of the initiative is to develop artificial intelligence capable of translating more than 200 world languages. So if you want to know more about the new Meta translation feature, just continue reading this article.
Read more: The technology present in financial institutions is a determining factor when choosing
see more
Low-wash: more and more people say goodbye to washing machines
Learn how to interact with Bard, Google's new chatbot and competitor…
The company's original concept did not support even half of today's languages, leaving millions of users unable to understand the content posted on the two largest social networks in the world (Facebook and Instagram). However, with the release of the new feature, these apps already include built-in translation engines, although they may not work as well in less common languages.
The new technology, known as NLLB-200, was developed to improve automatic translations on Facebook and Instagram, however it is too good to be used only in the company's products. So much so, that Meta AI researchers decided to publish the source code of the algorithm so that others can create solutions using it.
For example, there was no translation available for languages such as Zulu, Guarani and dialects spoken by African tribes, or at least, when it was available, the translation was inaccurate. In this sense, the lack of understanding creates a kind of access barrier to the contemporary world, dominated by social media, which further reinforces social inequality.
The new engine employs a different approach to ensure more focused results. Artificial intelligence includes a machine learning system that is fed a large volume of words. There are countless possible combinations, so the artificial intelligence compares the contents to determine which is the most suitable in that specific circumstance.
The challenge is to locate speakers of the language who are willing to “talk” extensively to the machine. In addition, there is also the problem of the absence of written grammatical rules in some languages, since certain languages were developed exclusively based on the spoken language.