Language learning is currently at the heart of Google's innovations. Through all of its tools - including the most advanced and "intelligent" - the tech giant hopes to be able to communicate with one and all.

Google recently unveiled the latest advances of its conversational AI service Bard, which is now based on the language model, PaLM 2. This is fed with texts from more than 100 languages, although, for the moment, Bard is only usable in English, Japanese and Korean. The aim is for it to be usable in about 40 languages by the end of the year.

But the Mountain View company does not intend to stop there, and recently said that it is working on an artificial intelligence model compatible with no less than 1,000 languages. Google has already announced that it has trained this language model on some 28 billion sentences covering more than 300 languages, the goal being to reach 1,000 soon. It can thus already automatically recognize, in audio, common languages such as English or Mandarin, but also rarer ones such as Azerbaijani.

Finally, during the Google I/O conference, the demonstration of a new, still experimental, technology called "Universal Translator" also caused a buzz. This is a video dubbing service based on artificial intelligence, capable of translating a speaker's voice and rendering it by matching their lip movements to the new language. Aware that this type of technology could very well be used for deepfakes, this tool will only be accessible to Google's partners, such as Arizona State University, and not to the public at large.

In addition to all these announcements and new features, Google has decided to integrate AI into its search engine, Gmail, Photos and Maps to help make them even better in any language.

 - AFP Relaxnews