The neural network learned to identify people from photos
Scientists from Russia taught a neural network to determine the nature of the people in the photo. Created by Russian specialists of the artificial intelligence performs the task 10% more accurate than humans.
A group of Russian experts from the Higher school of Economics under the leadership of Professor Eugene Osina presented a neural network that has learned to analyze the nature of the people with the help of physiognomy. This area is not considered a science and is only considered in theory, however, scientists have used it in teaching artificial intelligence as there is evidence that certain parameters are represented in it very accurately. To teach the algorithm to do the analysis, scientists used the services of 12 thousand volunteers. Each of them has provided their photo and was given a psychological test. The obtained data helped to train the neural network to determine the nature of the people in the pictures. When tested set, she was able to show a 60% probability of a match at 50% the average person and although it seems that a difference of 10% is not much, the experts insist on the opposite.
During the test, the scientists drew attention to some of the abilities of the AI. He took special care in identifying males of such traits as honesty and conscientiousness. Among the representatives of the beautiful gender neural network most accurately determined by the “extraversion” and “emotional stability.”