An “intelligent” system that draws conclusions about a person’s political orientation by simply analyzing photos of his face was developed by a researcher at Stanford University in the United States.

Dr. Michal Kozinski, who according to the British “Daily Mail” published the corresponding publication in “Nature”, had presented another controversial facial recognition algorithm in 2017, which – again – only uses photos to determine sexual orientation. The new system is essentially an extension of the previous algorithm in the area of ​​political beliefs.

The new machine learning algorithm was “trained” by analyzing over a million photos of people with known political beliefs from Facebook and other websites to learn facial expressions (expressing emotions, etc.) and other characteristics (zx the inclination of the Head) with the political preferences of each. One of Kozinski̵

7;s conclusions is that progressives are more likely to stare at the camera, while conservatives are more likely to look gross.

Kozinski claims that his system can “read” political orientation with a single photo with an accuracy of 70% to 72 %%, that is, it “occurs” seven out of ten times, much better than a random assessment (probability 50%) or five in ten times). According to him, the average person is about 55% correct when asked to visually assess how to position themselves in politics. The corresponding rates for “eye” recognition of sexual orientation are 76% (artificial intelligence) versus 56% (humans).

“People may overlook or misinterpret some clues, but that low level of accuracy isn’t necessarily the limit of what algorithms can do,” Kozinski said.

Source: ΑΠΕ-ΜΠΕ, newsbeezer.com