Silus Grok
Within the experiment, images of all 535 US congressmen were uploaded to the Rekognition system, and later compared to the public database of 25,000 photographs of previously detained persons. Neither of the congressmen was detained before, yet the system ranked 25 people as criminals. Of these, 39% are African American, which caused particular outrage at the American Civil Liberties Union. The list of the "criminals" mistakenly included John Lewis, a human rights activist and a member of the US House of Representatives.
“The system of recognizing faces and objects - no matter how accurate it is - can cost people freedom or even life. People with a different skin color are often subjected to harassment by the police, and Rekognition will only exacerbate this situation," the study said.
As an example, the organization cites a case that occurred in San Francisco in 2014. Then, an elderly African-American woman was handcuffed. Moreover, for no apparent reason, the police officers put her on her knees and pointed a gun at her. The reason for the arrest was the fact that the system mistakenly identified the license plate of her car with the number of a vehicle listed in the hijacking.
Amazon's press statement indicates that the test results could occur due to incorrect system settings. "During the experiment, the ACLU used a system with a default probability of 80% coincidence, but Amazon recommends law enforcers raise this parameter to 95%", they said.
Earlier, human rights defenders already expressed concern about sales of Amazon Rekognition to state structures. In addition, in May, African Americans in the United States Congress, which includes 48 African-American parliamentarians, sent a letter to Amazon head Jeffrey Bezos. In their appeal, they asked to stop the deliveries to the police system. In their opinion, its use will lead to abuses by the authorities, including mass surveillance of citizens. "We are concerned that this technology may have unintended negative consequences, which will affect African Americans, illegal immigrants and protesters," the statement reads.
Recall, the division of Amazon Web Services introduced Rekognition in 2016. The system uses artificial intelligence to quickly identify people, objects and texts in photos and videos. According to Amazon, the system is able to recognize up to 100 faces on images with a large number of people. Currently, the system is used by the police of Orlando (Florida) and the District of Washington (Oregon).
source: cnn.com
“The system of recognizing faces and objects - no matter how accurate it is - can cost people freedom or even life. People with a different skin color are often subjected to harassment by the police, and Rekognition will only exacerbate this situation," the study said.
As an example, the organization cites a case that occurred in San Francisco in 2014. Then, an elderly African-American woman was handcuffed. Moreover, for no apparent reason, the police officers put her on her knees and pointed a gun at her. The reason for the arrest was the fact that the system mistakenly identified the license plate of her car with the number of a vehicle listed in the hijacking.
Amazon's press statement indicates that the test results could occur due to incorrect system settings. "During the experiment, the ACLU used a system with a default probability of 80% coincidence, but Amazon recommends law enforcers raise this parameter to 95%", they said.
Earlier, human rights defenders already expressed concern about sales of Amazon Rekognition to state structures. In addition, in May, African Americans in the United States Congress, which includes 48 African-American parliamentarians, sent a letter to Amazon head Jeffrey Bezos. In their appeal, they asked to stop the deliveries to the police system. In their opinion, its use will lead to abuses by the authorities, including mass surveillance of citizens. "We are concerned that this technology may have unintended negative consequences, which will affect African Americans, illegal immigrants and protesters," the statement reads.
Recall, the division of Amazon Web Services introduced Rekognition in 2016. The system uses artificial intelligence to quickly identify people, objects and texts in photos and videos. According to Amazon, the system is able to recognize up to 100 faces on images with a large number of people. Currently, the system is used by the police of Orlando (Florida) and the District of Washington (Oregon).
source: cnn.com