More than 7,000 languages are currently spoken on this planet, and Meta apparently wants to understand them all. Six months ago, the company launched its ambitious No Language Left Behind (NLLB) project, which trains AI to seamlessly translate between numerous languages without having to go through English first. On Wednesday, the company announced its first major success. NLLB-200. It is an AI model that can speak 200 languages, including less commonly spoken languages such as Lao and Kamba in Asia and Africa.
The NLLB-200 can translate 55 African languages with “high-quality results,” according to a blog post by the company Wednesday. Meta says the model’s performance on the FLORES-101 benchmark outstrips current cutting-edge models by an average of 44 percent and for certain African and Indian dialects by up to 70 percent.
Translating between two specific languages – especially if neither is English – has proven to be a significant challenge for AI language models, as many of these translation systems rely in part on written data from the internet for education. It is very easy to do if you speak what this sentence is, much more difficult if you are looking for quality content on Fan or Kikuyu.
Like most other publicly promoted AI programs, Meta has decided to open source the NLLB-200 and provide nonprofits with a $200,000 grant to develop real-world applications for the technology. For example, apps like Facebook News Feed or Instagram. The meta post reads, “Imagine you visit a favorite Facebook group, stumble upon a post in Igbo or Luganda and can understand it in your language with just a click of a button.” You can get an idea of how the new model works in Meta. demo site.
All products recommended by Engadget are selected by our editorial team independent of our parent company. Some of our stories contain affiliate links. If you buy something through one of these links, we may earn an affiliate commission.