recently there have been are a lot of developments that have become a reality only through the use of neural networks. for example, researchers at the Massachusetts Institute of technology have developed an algorithm which is based on the image creates a short video about what, in his opinion, should occur in the future. at the heart of this technology are neural networks, machine learning and computer vision. The researchers “fed” the system about 2 million different videos for training, so she learned to predict the next point in the frame. for example, if the picture depicted a car, then the neural network will show the video, where it goes, and goes on
Interestingly, on the creation of a video work simultaneously two neural networks:. one directly causes moving objects in the frame, while the second determines whether the received video real or simulated. The result is considered successful if the system itself has failed to recognize that she is his and created. At the moment, this technology allows you to create video clips in length of only 1.5 seconds, and they do not seem too complex or realistic, but the waves of the sea, the movement of the grass or the human system can completely recreate.
If the scientists manage to teach technology to predict the future for more than 1.5 seconds, it can make cars more intelligent self-managed, as they will be able to predict the movement of pedestrians and other vehicles to avoid accidents. Also, this technology can lead to a revolution in video compression, as will no longer need to store all the frames completely, because the algorithm can recreate them at any time.