Various artificial intelligence systems already know how to do such things that it is time to think that the plot of the Terminator movie franchise is not such a fantasy. Moreover, every day AI is learning more and more new techniques. For example, Google recently trained its artificial intelligence to identify the voices of individuals in crowd noise with amazing accuracy.
To begin with, Google engineers created a machine learning-based technology that reveals the voices of individuals through the "party effect." This is what we all know how to do: communicating with a person surrounded by extraneous noises, we can understand what he is talking about, analyzing not only speech, but also facial expressions, filtering out “interfering” sounds. Google's AI works in a similar way: by analyzing the facial muscles and all the noise, it selects the voices of different people and creates separate audio tracks from them. Then Google trained artificial intelligence to recognize people's faces, isolate their voices and suppress extraneous noises.
Thus, the system can select the required person from the crowd of noisy people and carry out recognition of his voice in real time. This is possible thanks to the fact that Google engineers "coached" AI in more than 100,000 video lectures and conversational videos from YouTube with background noise. The new development, according to Google, can be used, for example, to improve the quality of video chats, so that during group communication, the user can focus on one interlocutor. You can watch the result of the Google algorithm in the video below.
Vladimir Kuznetsov