Advancements are made in AI models everyday in this age of technology and some have even started to believe that AI has become sentient. While the jury is still out on the validity of that claim, Meta has announced yet another language model that it claims is significantly better than the competition at translations.
The new AI model is called NLLB-200 where "NLLB" is an abbreviation for "No Language Left Behind" and the "200" refers to the number of languages it can translate. Meta has also open-sourced the model over at GitHub and it has multiple variants starting from 600 million parameters and going up to 54.5 billion parameters, depending upon your use-case.
Meta has boasted that NLLB-200 performed better at an average of 44% compared to other AI research in this area. In fact, when it comes to some Indian- and African-based languages, it was better by 70%. These results are based on the FLORES-200 dataset that Meta built for evaluation in 40,000 different language directions. The dataset is publicly available for developers to use too.
The company further announced that:
We're also awarding up to $200,000 of grants for impactful uses of NLLB-200 to researchers and nonprofit organizations with initiatives focused on sustainability, food security, gender-based violence, education or other areas in support of the UN Sustainable Development Goals. Nonprofits interested in using NLLB-200 to translate two or more African languages, as well as researchers working in linguistics, machine translation and language technology, are invited to apply.
Meta says that this model can result in improvements in multiple use-cases such as building digital assistants or mechanisms to add subtitles for movies. For its own part, it intends to use NLLB-200 across its social media platforms (Facebook, Instagram), and the development of the metaverse.
If NLLB-200 sounds interesting to you, you can check out the open-source GitHub repository, try the demo, and find out more details about the development process here and here.