Facebook’s parent company Meta is pushing the boundaries of AI technology and breaking down language barriers like never before. Meta’s new AI translation model — No-Language-Left-Behind 200 (NLLB-200) — has broken a translation milestone.
This single model can translate 200 different languages with detail and accuracy — a feat no other AI translation model has accomplished. NLLB-200 has significant implications for AI, translation, and the future of the internet.
Introducing Meta’s No-Language-Left-Behind 200
Meta’s new NLLB-200 AI model is the first of its kind to translate between 200 languages in a single model effectively. In the past, various models were typically necessary for different languages.
The company also included numerous languages that were either nonexistent in these algorithms or had poor translations due to limited training data. This helps make technology much more accessible to many people.
They prioritized inclusion for NLLB-200 and made it a point to create accurate and detailed translations for historically excluded or poorly translated languages. These frequently lacked good AI translation options due to a lack of text, data, and resources for the language like books and media.
“Low-resource” languages are more challenging to create translation models for because the AI has less data to learn from. Meta was able to overcome this challenge — they even went so far as to seek out native speakers of obscure languages when adequate resources were unavailable.
Meta will be open-sourcing the massive dataset it used to train NLLB-200 so more translation algorithms can be inclusive and accurate. They will use the model on their applications and websites — as well as Metaverse apps and experiential and educational media.
A Critical Step Forward for AI Translation and Communication
Meta’s NLLB-200 translation algorithm is more than just another AI translator. This model is a groundbreaking step forward for AI and communication.
For one thing, NLLB-200 will help expand communication capabilities to people worldwide, regardless of what language they speak. Even in an era where space travel and virtual reality are commonly used technologies, basic communication capabilities remain out of reach for billions of people. In the U.S., 39% of people don’t have broadband internet access in rural areas.
Meta is working to bridge gaps like this, expanding access to the opportunities the internet and emerging technologies have to offer. For example, Meta is working hard to develop VR technology’s availability to more users.
In 2021, they offered free Quest 2 headsets to all Facebook employees. Meta even announced in 2022 they would no longer require a Facebook account to use their Quest VR headsets. Eliminating a rule like this further expands how many people can use the technology.
The NLLB-200 translation algorithm plays a vital role in Meta’s accessibility initiatives. One of the top priorities for the project is an “inclusive Metaverse”, clearly indicating that Meta plans to apply NLLB-200 to their expanding VR platform.
While the Metaverse hasn’t become ubiquitous yet, many believe that VR is the next significant technological shift — the last one being the smartphone. If Meta does want to lead the emerging VR industry, a translation algorithm like NLLB-200 can ensure the Metaverse is accessible to everyone.
This is already a problem on the Internet as it is. Valuable sources of free information like Wikipedia are not available in many languages, leaving people who don’t speak English at a disadvantage. Free learning resources or career assistance on the internet is not accessible to billions of people simply due to language barriers.
Similarly, people everywhere could appreciate a wealth of cultural and recreational media like recipes and cookbooks if something could translate them effectively. Meta designed the NLLB-200 algorithm to help address these language barrier issues and future ones, such as language barriers in VR.
Advancing Language for an Inclusive Future
Meta is helping to build a more inclusive future and internet with the No-Language-Left-Behind-200 translation AI. While there are still plenty of languages to add for greater accessibility, this is a significant first step.
With the NLLB-200, Meta will help to expand the accessibility of content all over the Internet. The model will also help make emerging mainstream technologies like VR and the Metaverse accessible to everyone, no matter what a person’s native language is.
April Miller is a staff writer at ReHack Magazine who specializes in AI, machine learning while writing on topics across the technology sphere. You can find her work on ReHack.com and by following ReHack's Twitter page.