Meta has announced the introduction of a single AI model capable of translating over 200 languages, including 55 African languages. The open-source model, named ‘No Language Left Behind’ NLLB-200, is the first of its type.
The single AI model was created with high-quality translation skills in order to serve some of the world’s low-resource languages, as there is insufficient data or human translations to enable AI models to learn these languages, particularly African languages.
In a Facebook post made by Meta CEO Mark Zuckerberg, he expressed his satisfaction with this new milestone while also explaining that the team and expert translators worked hard to make this project a success.
He also stated that Meta’s goal with this initiative is to redefine the narrative of language problems and exclusivity on the internet by incorporating additional machine translation systems to accommodate more languages.
To give a sense of the scale, the 200-language model has over 50 billion parameters, and we trained it using our new Research SuperCluster, which is one of the world’s fastest AI supercomputers. The advances here will enable more than 25 billion translations every day across our apps. Communicating across languages is one superpower that AI provides, but as we keep advancing our AI work it’s improving everything we do — from showing the most interesting content on Facebook and Instagram, to recommending more relevant ads, to keeping our services safe for everyone.Mark Zuckerberg, Meta Chief Executive Officer
Meta intends to integrate its AI initiative in other social media apps such as Facebook and Instagram to help improve and expand language translations.
Language barrier in Africa
Language barriers have long been a source of contention among individuals, and this is evident even in the digital age. Despite the fact that there have been hundreds of high-quality translation tools, most of these tools do not cover a wider range of languages, particularly low-resource languages.
And, with the evolution of the digital world, billions of people today find it difficult to participate in online conversations or access digital content in their preferred or native languages, particularly in Africa.
There are over 2000 languages spoken in Africa alone, making it one of the most linguistically diverse continents, and in this part of the world, the language barrier is unavoidable. With the inclusion of 55 African languages in the machine translation research model, Meta intends to offer a sense of belonging among individuals, particularly on their platforms.
Balkissa Ide Siddo, Public Policy Director for Africa, believes that this research will not only provide people with better access to content in their preferred languages, but will also promote more accurate and precise information, which will help curb harmful content, misinformation, election dishonesty, and instances of online sexual exploitation and human trafficking.
Meta AI plans
In a bid to confirm its accuracy, Meta produced a new evaluation dataset, FLORES-200, to measure the AI model, NLLB-200‘s performance in each language, and the results showed that NLLB-200 does actually outperform the prior state of the art by 44 per cent.
The corporation is so pleased with the outcome of this research project that it has decided to open-source the AI model along with other research tools to assist other researchers in developing further technologies that will incorporate far more languages.
Meta has also partnered with the Wikimedia Foundation, a non-profit organisation that hosts Wikipedia and other free knowledge projects, and the NLLB research’s model-based approaches are currently being adapted to translation systems used by Wikipedia editors.
Meta AI is currently offering fund of up to $200,000 to non-profit organisations for NLLB-200 real-world applications.
Get the best of Africa’s daily tech to your inbox – first thing every morning.
Join the community now!