0

Meta is working on a special translator – learning the language is no longer necessary

Share
  • February 24, 2022

Mark Zuckerberg wants to create an artificial intelligence that can translate any language in the world. This is supposed to be the next important step in the process of building a meta-universe.

Meta has set a very ambitious goal – to create a new three-dimensional Internet – a meta-universe. Will they be able to build a virtual world straight out of sci-fi movies? No one knows, but one thing is certain. By investing huge money in the realization of such a distant goal, Mark Zuckerberg’s company will certainly contribute to the development of many new technologies. One of these will be artificial intelligence, and that is what the last Meta conference was devoted to.

The biggest star in the presentation of the American corporation were two projects related to the translation of human speech:

No Lenguage Left Behind, an artificial intelligence model capable of learning any human language;

Universal Speech Translator – an instant speech translator that supports all the languages of the world.

Both projects are related to each other and are based on an artificial intelligence machine learning method that Meta is intensively developing. It is about self-monitored learning, in which a machine learns to recognize, distinguish and understand received data on its own. This makes it possible to use untagged data for training, which does not require prior correct classification by humans. Thus, Meta plans to create a translator that supports even less popular languages that do not exist in text form.

But that’s not all. Today, most translators first translate each language into English and then into the target language. This is due to many problems. With self-guided learning, Meta wants to create a model that translates all of the world’s languages without using English.