SAN FRANCISCO – Controversial Amazon. Englisch: / en / movies / oliver – twist Test of the American Civil Liberties Union program, the civil rights group said Thursday

During their test, the ACLU scanned photos of all members of Congress and let the system compare them to a public database of 25,000 mugshots.

The group used the standard "trust Thresh old" setting of 80 percent for Rekognition, meaning the test counted a visual field of 80 percent or more.

Against this background, the system identified 28 members of Congress, one of whom was a disproportionate number of people of color, and instead identified them as totally different individuals arrested for committing a crime.

The faces of congressmen in the test are Republicans and Democrats, men and women and legislators of all ages.

Amazon responded that when using face recognition for law enforcement activities, it is recommended to set the confidence threshold to 95 percent or higher.

A spokesperson for Amazon Web Services said in a statement that the test results could be improved by raising the confidence threshold. While 80 percent represent an acceptable threshold for photographs of everyday objects and objects, it is not appropriate to identify individuals with "reasonable assurance".

In their report on the results, the ACLU said the default setting for the program was 80 percent and the Amazon recommends this level for face-based user verification.

The tool is used for face detection in arenas outside law enforcement. During the royal wedding of Prince Harry and Meghan Markle in May, British broadcaster Sky News used Rekognition to identify celebrities as they entered Windsor Castle.

The software was also used by Pinterest to tailor images to stores tracking people, identifying potentially insecure or inappropriate content online, and finding text in images.

Data protection and policing

Amazon has recently come under fire for selling face recognition services to law enforcement agencies in pursuit of persecuting people who go about their daily lives, or in political protests or other situations Most people assume that they are anonymous.

These concerns have prompted civil rights groups, privacy advocates, and even some of Amazon's CEO and CEO Jeff Bezos to ban police and federal agencies from using face recognition technology.

The results of the ACLU test "demonstrate why Congress should join the ACLU in calling for a moratorium on the use of law enforcement surveillance," wrote Jacob Snow, a lawyer for technology and civil rights for the ACLU in Northern California.

Two years ago, Amazon developed the Face and Product Recognition Tool so customers can quickly search a database of images and search for matches. Recognition requires the user to have two sets of pictures. The first is generally a large database of known people. The user then sends images to individual who then compares the software with those in the large database to find what they think matches are.

Snow said the product had been sold "aggressively" to the police. At least two agencies, one in Orlando, Florida, and one in Washington County, Oregon, are currently testing recognition.

The Washington County program analyst says it would never rely on face recognition software to talk to, let alone arrest, a potential suspect.

The department does not set a confidence threshold at all, because all decisions are made by humans, said Chris Adzima, the chief Information Systems analyst with the Washington County Sheriff's Office in Hillsboro, Oregon.

"If we have a picture of an active investigation, the investigator will put it into our system and the system will spit out the five most likely outcomes, then the investigator will look through those five to see if any of these are possible hints" , he said.

Even then, the investigator must do the proper due diligence and run the name to see if the person had a record or was known to contact potential victims

While Adzima said that Washington County was successful with facial recognition "In order to identify people who are ultimately committed to crime, almost none of them had a confidence threshold of 95 percent," he said. [19659006Thefacerecognitiontechnologywassuccessfullyarrestedbythemanidentifyinghimselfforshootingin Gazette newsroom in Baltimore, Maryland.

More: Amazon Should Stop Selling Facial Recognition Software to Police, ACLU, and Other Rights Groups

More: Orlando police decide the controversy Continue Testing Amazon Face Detection Program

But the ACLU and other privacy advocates say technology is a violation of privacy. And they say it could be used to target and persecute immigrants or demonstrators.

In May, 34 civil rights groups sent a letter to Amazon boss Jeff Bezos, saying that "people should be free to walk down the street without being watched by the government."

The same month members of the Congressional Black Caucus also sent a letter to Bezos, saying that they were affected by the "profound negative unintended consequences" this technology would have for African Americans, undocumented immigrants and demonstrators.

"The race-based" blind spots "in artificial intelligence, especially those manifested in facial recognition technology, have been well documented," the letter says.

These "blind spots" in facial recognition AI include an incident in 2015 in which a Google photo application identified images of African American users as "gorillas", and a study published earlier this year by the Massachusetts Institute of Technology Face recognition software was published, used to identify a person's gender, had an error rate of 0.8 percent for fair-skinned men and 34.7 percent for dark-skinned women. The study used three different types of commercial facial recognition software.

Six members of the Congressional Black Caucus were misidentified in the ACLU test of recognition.

repetitions. Jimmy Gomez (D-Calif.) And John Lewis (D-Ga.), Who were both misidentified during the test, sent a letter to Bezos on Thursday asking them to meet immediately to discuss the "failures" of the technology to address in order to avoid inaccurate results.

Read or Share This Story: