Meta, the parent company of Facebook, has produced an AI that can translate 200 languages in real time.
Six months ago, the company launched the No Language Left Behind (NLLB) project, aimed at training an AI to translate between languages that have commonly been overlooked by such tech companies.
English, Mandarin, Spanish and Arabic are the four main languages that dominate the web, leaving many people without web content in their native tongue. In particular, the NLLB project aimed to address the “hundreds of millions of people who speak the many languages of Africa and Asia”.
Now the company has announced an important breakthrough in the shape of NLLB-200, an AI model that can translate between 200 languages far more accurately than was previously possible. This includes so-called “low-resource languages” such as Asturian, Luganda, and Urdu.
Meta claims that its new AI model can translate 55 African languages with “high-quality results”.
Meta calculates that NLLB-200 can translate with a 44% higher degree of accuracy than previous standards, and 70% more in the case of some African and Indian languages.
This being Meta, all of its breakthroughs have to have at least one eye on the emerging metaverse. To that end, the company claims that “integrating real-time AR/VR text translation in hundreds of languages is a priority” as it builds out its ambitious virtual universe.
Meta has announced that it is making the NLLB-200 model open source so developers can work on incorporating it and researchers can extend its reach to even more languages. It also intends to supply $200,000 worth of grants to nonprofit organisations for “real world applications for NLLB-200”.