Meta recently emerged as the owner of  Facebook, Instagram, and WhatsApp. The company’s rapidly growing reach stressed it forward in removing barriers separating individuals from connecting worldwide. 

The main barrier comes up as a language barrier, and to remove this, Meta announced an ambitious new AI research project, where they will create a translation software that works for “everyone in the world.” 

Meta announced this project as part of an event where the major objective was to discuss the broad range of benefits that AI can offer the company’s metaverse plans according to the company’s beliefs. 

Mark Zuckerberg, CEO of Meta, said that from the list of superpowers people dreamed of forever, one is the ability to communicate with anyone in any language, and bringing this dream into real AI will help.

Meta Universal Speech Translator: Insight 

According to Meta, the most common languages spoken are English, Spanish, and Mandarin, and the current translation tools cater to these languages only. Still, roughly around 20 percent of the total world’s population speak different languages that these systems do not cover. 

Meta says these under-served languages are hard to implement in AI and to overcome these challenges; they plan to work on their language translator project. This project will deploy new machine learning techniques targeting two specific areas these are- 

1. The first one is No Language Left Behind. This will mainly concentrate on building advanced language translating AI models that need short training examples to learn language translation. 
2. The second one is the Universal Speech Translator. This will aim to build systems that work in real-time and directly translate speech from one language to another without involving written components like other translating apps. 

So far, about the timeframe, Meta did not mention the exact time duration this project will take to complete. Also, nothing is mentioned about the roadmap for major milestones they will achieve until they reach their ultimate goal. Instead, Meta stressed the Universal Language Translation’s utopian possibilities.

To be more specific with the company’s exact words in their released blog post says: Apart from eliminating the language barriers and making information access in the native language possible for billions the company will also look for advanced machine translation tools that will fundamentally change the way people connect with the world and share their ideas.

Moreover, this new technology will benefit Meta for its globe-spanning products and stretching its reach. This technology will very soon become an essential communication tool for millions of users from all across the globe. 

Meta technically believes that the upcoming universal translation software can be a VR and AR killer that Meta is already building. But Meta has good business sense and knows how to bring tools that can benefit all. 

AI Translation Comes Up With Number Of Hidden Problems

In the past few years, advances in machine learning have improved a lot and are still gaining pace. The major improvement is made in machine translation speed and accuracy to achieve the ultimate goal of translating direct speech in real-time.

Many big names like Google & Apple already offer their users free AI translation tools. These tools are beneficial for global interaction and tourism, and, apart from this, it also provides incalculable benefits in different sectors. 

But the underlying translating technology has some basic problems, where critics notice that the machine translation misses nuances that are very critical for human speakers. They also notice that it injects gendered bias into its outputs and sometimes throws weird and unexpected errors that only a computer can throw. 

Some speakers also say that when big tech controls the ability to translate their words for uncommon languages, they will lose hold of their speech and culture. When platforms like Facebook and Instagram automatically apply this translocation tool, such errors get critical and a master consideration.

For example, in 2017 we saw a case where Israeli police arrested a Palestinian man for offending Israeli law by posting a piece of content on Facebook. Later, after the investigation, it was found that the machine translation software on Facebook mistranslated the complete post he shared. 

He wrote “good morning” in Arabic, but the translation software of Facebook translated it to  “hurt them” in English, and in Hebrew, it got translated to “attack them.”

Data Scarcity: Biggest Obstacle For Meta

When it comes to building translation tools, Data scarcity comes up as the biggest obstacle for Meta. The systems we use for text translations currently work on different annotated data and from millions of sentences. This is the reason why only a few languages that dominate the web are available for translation, and the rest are unavailable. 

Expanding the list of languages also expands the need to find different ways to acquire and use training examples from languages having a sparse web presence. 

But, even after all this, Meta sounds confident in its aim, and we will soon encounter this AI-based universal speech translator in early 2023. 

Conclusion 

This AI project by Meta will streamline users’ interactions with people who speak different languages, especially those languages which are not common to us. Universal Translator can play crucial roles for different social media platforms where everyone wants to increase their reach globally without losing their identity. 

We discussed everything that Meta revealed in their event and blog post. Every big invention needs time to come alive, and the same goes for this advanced AI project. Till then, you can check out our different blogs.

Also Read: Why Is Facebook Losing Its Touch?