[ad_1]
The Metaverse is an emotive topic – people fall into the categories of listeners, skeptics, and cynics.
Tech companies hope it will be the next big thing. Heck, Facebook even changed its name to Meta in anticipation of the new technology, and invested over $15 billion to build this new virtual world.
But the bet is not viable. At least, not now. Meta’s valuation has dropped by $700 billion, the AR/VR headsets needed to enter the Metaverse are still not distributed, and investors are saying they’re not seeing the value in the concept.
So far, you could say the Metaverse project has been a flop. Except for one area – the development of small language technology. Thanks to the Metaverse commitment, Google, Nvidia and Metaverse themselves are bridging the small-language technology gap by using AI for speech-to-speech (S2S) translation.
This is good news for small countries with obscure languages that are not high on the priority list for language technicians like Google Translate. And while there have been previous efforts to develop translation technologies for smaller languages, the (hoped-for) rise of Metaverse has catapulted them into prominence.
Metaverse for small language AI technology
Metaverse is a shared virtual/digital world concept where you can interact with the environment and interact with other people through VR/AR headsets. The site is completely virtual, and like the Internet, it has no borders. And therein lies the problem. If people who speak different languages meet, they cannot communicate with each other.
Big tech giants have focused on developing technology to translate large languages, developed algorithms and tried to apply those principles easily to all languages. That makes for a great experience for large language speakers, but when you start adding smaller languages into the mix, translation accuracy drops dramatically. This creates a difference in the availability of technology products for minority language speakers. For example, Siri does not speak Latvian. Language technologies take longer to reach minority speakers.
But that’s about to change thanks to Metaverse. Since one of the key requirements for achieving adoption is the elimination of language barriers in the Metaverse, the participating parties (most likely to build from and/or benefit from the Metaverse) are interested in overcoming this problem.
For this reason, Google, Meta and Nvidia have plans to develop technologies that support small languages in the S2S translation.
Are the tech giants late to the small language AI game?
In the past month, Google has announced the launch of a translation center for document translation and an AI-driven universal speech model. Meta also announced a Universal Speech Translator (UST), which can handle both written and unwritten languages. NVD quickly followed up on their own announcement – a speech AI ecosystem to accelerate the development of automatic speech recognition models.
But are you late to the game?
Tilde CEO Andres Vasiljeves said these actions are long overdue. Small languages have been neglected by tech giants for decades. To meet the needs of these communities, Tilde has been making the best of AI-based speech and translation technologies for over a decade. Not only that, they beat all other MT service providers in WMT competitions – such as the Machine Translation Olympics – for three years in a row in the “small languages” category with the highest accuracy.
How did a small company from Latvia beat the world’s leading provider of translation technology? First by developing AI technology for some of the world’s most grammatically-complex languages and then applying those algorithms to other languages.
“Google and other companies focus on big languages first, and then try to apply these algorithms to small languages. This doesn’t work effectively. What we did instead, is start with the complex languages, which require a different approach to train AI. In English, maybe 500k-700k words.” You only have forms. For example, in Latvian you have over 22 million possible word forms. Using these techniques, we can successfully use AI in other languages like Estonian and Finnish. – Andres Vasiljeves
Tilde AI has already been integrated into Microsoft products, and has also become the official translation tool of the Finnish Prime Minister’s Office. The company welcomes the news that big players are paying more attention to small languages, but doubts their ability to do well.
“It’s one thing to announce that you’re doing it, but until you check the accuracy of the translations, you can’t guarantee the success of the effort. Additionally, one-size-fits-all definitions don’t always fit everyone. Our high level of accuracy depends not only on the ability to train AI in a small language, but also for a specific industry – for example, legal affairs or construction. As with human translators, it’s not enough to be good at the language, you also have to be a subject matter expert.”
Although recent announcements seem promising, Vasiljevs says universal, high-precision AI in small languages is still out of reach for even the most well-funded corporations.
Enabling technology at the expense of privacy?
One aspect Vasiljeves highlights is that freemium products like Google and Meta offer their services in exchange for the right to do whatever they want with their user data. In an age of privacy concerns, it’s worth taking a moment to think about what services you’re using, what data you’re processing, who has access to that data, and whether you’re okay with it.
This is a valid point when analyzing the Metaverse, where every spoken word becomes data processed by whoever provides the translation engine, assuming their terms of service grant access to that data. Do we, then, want to put every word of speech (in the metaverse) into the hands of corporations with less than stellar reputations for data management? In the year With an estimated 8.4 billion voice-activated assistants in devices by 2024, the world seems poised to hand over their voice data.
In general, the inevitable growth of the metaverse seems to be driving the demand for minimal language technology. And while we don’t know what the future holds in terms of living our lives in virtual reality, we do know that communication between us will be a little easier.
About the author
Julia Gifford is the founder of Truesix, an award-winning content marketing and PR agency focused on visibility for high-tech organizations. She was a member of the founding team of Latvia’s first Unicorn Publishing Group. She is an editor and contributor at Latvian lab and tech.u, one of Europe’s leading news platforms.
Featured Image: ©agungkreatif
[ad_2]
Source link
