The British division of Google just announced that, after using hundreds of hours of videos and programming from the BBC, namely nothing less than 5,000 hours of video which included some 110,000 different phrases and 17,500 different words, have managed to make DeepMind software able to read lips, thus fulfilling his aim of getting your software to be able to understand human language without it both sound much better.
Now, as comment makers own this project, at the moment are not talking about how DeepMind has managed to reach perfection but the truth is that, after numerous tests, if that have come to the conclusion that, with 46.8% of words read correctly is much more effective than a human who is dedicated to this type of translations in a professional manner that could hit 12.14% of occasions.
After many hours of training, DeepMind dramatically increases your ability to read the lips of a user.
Thanks to the development of experiments and projects as which have led to out them boys of Google, within little time can see how evolve the systems intended to people with disabilities hearing. One of the closest applications that can have this project would be, as it is just to announce, to achieve a better communication between users and the virtual assistants such as Siri or Cortana visually.
