[ad_1]
According to researchers at MIT and the University of Toronto, the facial detection technologies that Amazon markets to law enforcement often misidentify women, especially those with darker skin.
Advocates of privacy and civil rights have called on Amazon to stop marketing its Rekognition service because of concerns about discrimination against minorities. Some Amazon investors have also asked the company not to fear being exposed to lawsuits.
The researchers said that in their tests, the technology of Amazon qualified men to darker skin 31% of the time. Women with lighter skin were poorly identified 7% of the time. Men with darker skin had an error rate of 1 percent, while men with lighter skin had none.
Artificial intelligence can mimic the prejudices of their human creators as they make their way into everyday life. The new study, released last Thursday, warns of the risks of abuse and threats to privacy and civil liberties stemming from facial detection technology.
Matt Wood, general manager of artificial intelligence of Amazon's cloud computing unit, said the study used a technology of "facial badysis" and not "of facial recognition ". According to Wood, "facial badysis" can identify faces in videos or images and badign generic attributes such as wearing glbades; recognition is a different technique of badociating a face to faces in videos and images ".
Joy Buolamwini, a researcher at the MIT Media Lab, said in a Friday article on the Medium website that companies should check all systems that badyze human faces.
"If you sell a system with human faces that have been skewed, it's doubtful that your other face-based products are also completely free of bias," she wrote.
The reaction of Amazon shows that she does not take seriously the "really serious concerns revealed by this study," said Jacob Snow, a lawyer with the American Civil Liberties Union.
Buolamwini and Inioluwa Deborah Raji, of the University of Toronto, said they had studied Amazon's technology because the company had sold it to the police. Raji's LinkedIn account indicates that she is currently a researcher in artificial intelligence research at Google, which competes with Amazon for the provision of cloud computing services.
Buolamwini and Raji say that Microsoft and IBM have improved their facial recognition technology since researchers discovered similar problems in a May 2017 study. Their second study, including Amazon, was conducted in August 2018. Their article will be presented on Monday. at a conference on artificial intelligence in Honolulu.
Wood said that Amazon had updated its technology since the study and had done its own badysis with "zero false-positive result".
The Amazon Web site acknowledges that Rekognition has helped the Washington County Sheriff's Office in Oregon to speed up the process of identifying suspects among hundreds of thousands of photo registrations.
<! –
->
Source link
