Pages

Meta’s Open Source AI Model Can Swiftly Translate Up To 200 Different Languages

Meta is taking great pride in its latest open-source AI Model. The company says its ‘No Language Left Behind’ project has the capability to swiftly translate up to 200 different languages. And let’s not forget how it's recently been declared open-sourced for maximal benefit.

The model was first revealed in February of this year. And that’s when the company shed light on how it was working on designing a new AI-based model which could learn a series of different languages without the need for an extensive number of examples, from which it could receive training through.

The whole idea behind the revolutionary and innovative project had to do with users benefitting from hundreds of various language translations. Be it Urdu and French to Asturian and more- the options are truly unlimited.

Hence, that’s why the venture was soon dubbed NLLB-20 because truly, Meta does not plan on leaving any specific language behind.

The translation quality was even compared by some researchers to previous ones obtained with the help of AI technology. And that’s when it was revealed how Meta’s model managed to obtain accuracy that was almost 44% greater than all others. And as you can tell, that’s a wonderful result and one that Meta is definitely proud of.

But wait, the results got better when some translations related to African or even Indian dialects were compared. Here is where accuracy of more than 70% was observed.

Based on these findings, the tech giant says improved translations may end up providing more support for billions of translations that are observed on a daily basis across the Facebook app, Instagram, and other similar social networks.

In addition to that, the new model will reportedly be able to highlight content that’s explicit, harmful, and ends up turning into misinformation too. Common examples of issues that Meta plans on curbing include matters related to sexual abuse, human trafficking, and controversial election campaigns.

Topics centered around these genres usually end up producing false reports and hence account for a great percentage of misinformation spread across the app.

Meta also says it hopes the new AI model would assist in allowing the advancements of other similar technologies. For instance, you could soon see the arrival of systems generating better or more suitable subtitles. Similarly, we could see the arrival of building assistants as well.

In addition to that, we’re hearing more news on how Meta plans on establishing a great partnership with Wikimedia. This way, the AI-based model would also assist in the translations of numerous articles present across Wikipedia.

And how can we forget about the Metaverse? Meta hopes the language-themed model would be a key helper in making the virtual experience more accessible to different users around the globe. But only if they can bear the expenses related to VR headsets.

As of now, users can get a feel of how the AI model actually works by visiting the company’s website. Here, stories are available in diverse languages which are up for translation into any language as per your liking.


Read next: Facebook In Trouble As Ex-Employee Says The App Gave Its Staff Full Access To Deleted User Data

No comments: