Party officials came to the police to sell software. The company offers a free trial and a minimum annual fee of $ 2,000.
According to Hotan, Schwartz used his political connections to help government officials discover the face recognition system. “I am very pleased to have the opportunity to help Huo Tang build Mingshi as a mission-driven organization, help law enforcement to protect children, and enhance the safety of communities across the country,” Schwartz said through a spokesperson.
The company’s main contact with clients is Jessica Medeiros Garrison, who has participated in the Alabama Attorney General’s Republican campaign for Luther Strong. Related sources said that Garrison was the company’s “growth consultant.” Mingshi said this was a short-lived, unpaid role, and the company also invited Democrats to help sell its products.
The company’s most effective sales technique is to provide police with a 30-day free trial, and then police officers encourage their purchasing department to register the software and praise the tool to police officers in other police departments at meetings and online. Step by step, Huo Tang’s face recognition caused a sensation in the police circle.
In July last year, a police officer in Clifton, New Jersey, urged his superiors to purchase the software in an email because it “can identify a suspect within seconds.”
During the free trial of the department, Mingde identified the shoplifter, the Apple shop thief, and a Samaritan who acted with righteousness, who beat a gangster who threatened others with a knife.
The police officer wrote in an email, “You can use the telephoto lens to take photos secretly and enter them into the software without using police tracking operations.”
According to sales presentation materials from Mingshi, the application helps identify various groups of people, such as a person accused of sexual abuse of children, whose face appears in the mirror of someone’s gym photo; a series of mailbox thefts in Atlanta Behind the scenes is a black man; an anonymous person found dead on the sidewalk in Alabama, and a suspect in a bank multiple identity fraud case.
In Gainesville, Florida, police officer Nick Ferrara heard about Vision Corps in an advertisement facing financial crime investigators last summer. He said he previously relied only on FACES, a facial recognition tool provided by the Florida state government, which extracts face information from more than 30 million facial photos and photos from the automotive sector in Florida.
Ferrara finds that Mingshi ’s applications are more advanced, he said. Its national face image database is much larger. In addition to FACES, Mingshi ’s algorithm does not require the face of the camera to be photographed.
Sergeant Ferrara said, “With Vision, you can use it to take imperfect photos. A person can wear a hat or glasses, or a side view or a partial view of their face. “
He uploaded his photo to the system, which brought up his profile page on Venmo, a social payment website. He imported photos of old, unresolved cases and identified more than 30 suspects. In September last year, the Gainesville Police Department paid $ 10,000 for a one-year license for the company ’s face recognition system.
According to the company and government officials, federal law enforcement agencies, including the FBI and the Department of Homeland Security, as well as Canadian law enforcement authorities, are trying out software that is clearly seen.
Despite its growing popularity, Mesquite has avoided being mentioned publicly until late 2019, a Florida prosecutor has accused a woman of stealing two grills and a vacuum at Ace Hardware Store in Clermont Theft was found behind a vacuum cleaner. She was recognized when police pulled a still image from surveillance footage through Mingshi and brought it to her Facebook page. According to documents in the case, her identity was confirmed by tattoos visible in surveillance videos and Facebook photos.
“We’re all done”
However, Huo Tang also pointed out that the face recognition system is not effective at all times. Most of the photos in Mingshi’s database were taken at eye level. Most of the photos or videos uploaded by the police come from surveillance cameras mounted on the ceiling or wall.
“They put the surveillance camera too high,” Huo Tang sighed. “This angle is wrong for face recognition to work properly.”
The company says that despite this, its tool can find matching people 75% of the time. However, it is unclear how often the tool will mismatch because it has not been tested by an independent agency, such as the National Institute of Standards and Technology (a federal agency that evaluates the performance of face recognition algorithms) .
“We don’t have data to show that the company’s identification system is accurate,” said Claire Garvey, a researcher at the Georgetown University Privacy and Technology Center, who has studied government use of face recognition. “The larger the database, the greater the risk of misidentification due to the so-called ‘duplex body’ effect. They use a face database of a large number of random people found on the Internet.”
But current and former law enforcement officials say the identification software is effective. “For us, the trial process is mostly to see if it works,” said Cohen, a former Indiana police captain.
One of the reasons that Mingshi’s company identification system is popular is that its services are unique. That’s because Facebook and other social media sites forbid people from capturing images of users-Mingshi Company violates the site’s terms of service.
“A lot of people are doing this,” Huo Tang shrugged. “Facebook knows.”
Facebook spokesperson Jay Nancarrow said,And computer science professor Woodrow Hatzog believes that Vision is the latest evidence that the United States needs to ban face recognition.
Hartzog said, “We have been relying on industry efforts to come under our restraint and not accept such high-risk technologies, but now the dam is collapsing because of so much money on the table. I don’t see how we can prevent it in the future. The advantages of using face recognition technology in the context of surveillance abuse, and the only way to stop it is to ban it. “
Everyone knows your name
Huo Tang demonstrated the application in an interview at a Mingshi corporate office in Chelsea, Manhattan. He took a selfie and uploaded it to the system, and the application called up 23 of his photos. One was that he was naked, smoking a cigarette, and was covered in blood.
Huo Tang then took pictures of reporters, and the recognition system returned a lot of photo results, dating back to ten years ago, including photos that reporters have never seen before. When investigative journalists covered their noses and the bottom of their faces, the app still returned seven correct matches.
Police officers and investors at Mingshi predict that the company’s applications will eventually be open to the public.
Huo Tang said he was unwilling, “There will always be a group of bad people abusing it,” he said.
Even if Mingshi does not disclose its application software, other companies can imitate it, and the taboo is now broken.
Today, searching for a person’s photo is as easy as searching for a name. A stranger can eavesdrop on sensitive conversations, take pictures of the talker, and grasp personal secrets. In addition, people walking on the street will be recognized quickly, and it only takes a few clicks to get his home address, which will foreshadow The end of public privacy.
When asked what it means to bring such a power into this world, Huo Tang seemed surprised. “I have to think about it,” he said. “We think this is the best use of this technology.” (Tencent Technology Review / Cheng Xi)