Scientists have come up with a way to wean AI from racism

In late January, Amazon was again inthe center of the scandal associated with its artificial intelligence for face recognition. It turned out that its system Rekognition is not able to distinguish the gender of black people, and therefore it is much inferior in capabilities to analogues from Microsoft and other companies. The logical way to solve the problem would be to use better training data, but researchers from the laboratory of artificial intelligence at the University of Massachusetts came up with a more effective way.

It became known that they are developing an algorithmwhich automatically relieves the training data sets from photographs that can lead to an error in recognizing people with dark skin. The researchers do not promise that this method will give a 100% result, but the number of errors will clearly become less. This is evidenced by the internal tests of scientists: artificial intelligence has become 60% less biased, and at the same time has not lost its recognition accuracy. The technology will also reduce learning time - the use of large amounts of data has greatly slowed down the process.

Lack of bias - very importantcharacteristic for artificial intelligence. It is increasingly being used by the police, so even a small algorithm error can cause a completely innocent person to go to jail. Many people do not like the widespread use of facial recognition systems, but if they are to be used, it is important that they work as precisely as possible.

They are already used even in Russia. For example, during the 2018 World Cup, the face recognition algorithm was introduced into the Moscow Frunzenskaya metro station. Thanks to her, the police was able to quickly find the person who is wanted for robbery. He became the first criminal to be caught using a face recognition system. Its cost, by the way, is estimated at 4 billion rubles.

Want to talk about facial recognition systems? Welcome to the comments or to our Telegram-chat, where hundreds of participants are always happy to talk on the topic of science and technology!