Meta Says Its AI Translator Can Figure Out Unwritten Languages

Meta CEO Mark Zuckerberg says the tool will be open-sourced so more people can access it.

Alix Langone Former Reporter
Alix is a former CNET Money staff writer. She also previously reported on retirement and investing for Money.com and was a staff writer at Time magazine. Her work has also appeared in various publications, such as Fortune, InStyle and Travel + Leisure, and she worked in social media and digital production at NBC Nightly News with Lester Holt and NY1. She graduated from the Craig Newmark Graduate School of Journalism at CUNY and Villanova University. When not checking Twitter, Alix likes to hike, play tennis and watch her neighbors' dogs. Now based in Los Angeles, Alix doesn't miss the New York City subway one bit.
Alix Langone
Meta logo on a phone with a map of Europe
Sarah Tew/CNET

Meta announced on Wednesday what it's calling the "first AI-powered speech translation system for an unwritten language," according to a blog post from the tech company.

The system successfully translated the Taiwanese language Hokkien, a primarily oral language spoken across Taiwan. The system was developed as part of Meta's artificial intelligence project called Universal Speech Translator, or UST.

"The translation system is the first milestone for Meta AI's Universal Speech Translator project, which focuses on developing AI systems that provide real-time speech-to-speech translation across all languages, even primarily spoken ones," Meta said in a statement. 

Most previous AI-powered efforts for translating languages have focused on written language, Meta says, but one of its goals is to translate more languages that lack a formal writing system. According to Meta, there are more than 7,000 living languages, but "over 40% are primarily oral and do not have a standard or widely known writing system like Hokkien."

In a video announcement posted to Facebook , Mark Zuckerberg said Meta would be open-sourcing the tool so people can use the new AI system to translate more languages.