Computer engineers have developed a training system that allows the computer to predict the future. It is based on a special theory of relativity. To date, tests have been conducted on images and video files, and the machine predicted the appearance of new frames.
Unlike humans, in predicting the future, the computer can not connect the causes and consequences. That is why engineers are implementing a method of machine learning to obtain future results. The concept of such training is that linked images and video frames are consistently shown. In its turn, the computer reveals the logic of such images and is able to model the future on the basis of what it sees. The disadvantage of such prediction is that the computer has the property to make mistakes and accumulate them over time.
British scientists have taken a different method. They developed an algorithm based on the special theory of relativity. The basis of the method consists of the theory of space-method and method of light cones. Some area located in space-time, the central part of which is located in any episodes, is called a light cone. Scientists have limited the areas of past and future specifically for the episode under study. The inner part of the cone contains a large number of points related to the given episode by a cause-effect relation. The points that are outside the bounded area are neglected.
Using this method, the scientists have reduced the number of predicted variants, which in turn has reduced the percentage of errors.
In the future, such methods can form the basis of autopilots for driving a car, which will allow modeling the road situation.