The artificial intelligence of Google develops its own internal language capable of identifying different concepts and meanings.
This internal language is used as a link for translations from other languages.
In September, Google officially introduced the system Google Machine Translation Neural which used deep learning to produce better translations between different languages. Initially, this system allowed translations between Chinese and English, gradually expanding to more than 103 languages Google Translate currently supports.
In the process of expansion, the Google Machine Translation Neural system uses a system known as Zero-Shot Translation , which facilitates the process of expansion into more than 100 languages supported by Google Translate.
With this new method, the Google Machine Translation Neural system can translate between two different languages without first establishing a direct connection between the two . Thus, if the system is able to translate between Castilian and English between English and Korean, you will also be able to translate between Castilian and Korean without having to “teach” how.
Neural Machine Google has developed its own internal language translations to accelerate
But the really special feature of this new system is that Google does not use Machine Translation Neural English as a “bridge” to translate between Castilian and Korean. Instead, neural networks system are able to make connections between concepts and words that are not related by their programmers.
The Google system, therefore, produces a language used internally to represent different concepts and then use it as a link between the different languages.A breakthrough in artificial intelligence that surprised even those responsible for Google.