Google Translate learned to translate live human speech

Google has demonstrated a new mode of operation for its Google Translate translation system. Now it will be able to process extensive speeches in real time, such as lectures, politicians or artists. The new function is based on the work of special artificial intelligence - the company is asked to be lenient about his mistakes while he learns.

There are two tough conditions for the new option to apply. First, a good quality internet connection is required as all the work is done on Google servers. Secondly, live speech received through a microphone is subject to processing. This serves as a kind of restriction on the commercial application of the technology - it will be possible to listen to a live or recorded lecture in a foreign language, without subtitles, but not to use an interpreter in cinemas.

While working, Google Translate will compare the words heard among themselves, take into account the context of the statements, intonation and intuitively place punctuation marks. The work is carried out in real time, with constant editing of the text, so incidents are inevitable. But Google promises to deal with them as they learn AI on real-world assignments. So far, an artificial translator has a minimal set of languages: English, German, Spanish and French. The date of opening the service for public access has not yet been determined.