When tech companies created Facial Recognition Systems that are increasingly increasing government surveillance and removing personal privacy, they may get help from an unexpected source: your face.
Companies, universities and government laboratories have used millions of images collected from a hodgepodge of online sources to develop technology. Now, researchers have created an online tool, Exposing, That lets people search many of these image collections for their old photographs.
The tool, which matches images from the Flickr online photo-sharing service, provides a window on the vast amount of data needed to build a wide variety of AI technologies from facial recognition to online.Chatter. “
“People need to feel that some of their most intimate moments have been weaponized,” said Liz O’Sullivan, technology director of the Monitoring Technology Oversight Project, a monitoring and civil rights group. He collaborated on Exposing.AI with a researcher and artist Adam Harvey in Berlin.
Systems that use artificial intelligence do not magically become smart. They learn by pinning patterns in figures generated by humans – photos, voice recordings, books, Wikipedia articles and all other types of content. Technology is getting better all the time, but it can be learned by humans Prejudice against women and minorities.
People may not know that they are contributing to AI education. For some, It’s a curiosity. For others, it is very scary. And this May be against the law. A law in Illinois in 2008, the Biometric Information Privacy Act, imposes financial penalties if residents’ face scans are used without their consent.
In 2006, Brett Geller, a documentary filmmaker from Victoria, British Columbia, then uploaded his honeymoon photos to Flickr, a popular service. Nearly 15 years later, using an early version of Exposing.AI provided by Mr. Harvey, he found that hundreds of those photographs had made their way into multiple data sets, using facial recognition systems around the world. Can be used to train.
Flickr, which was bought and sold by many companies over the years and is now owned by photo-sharing service Smaugmug, allowed users to share their photos under what is called a Creative Commons license. That license, which is common on Internet sites, means that some people may use photographs with certain restrictions, although these restrictions may have been ignored. In 2014, Flickr-owned Yahoo at the time used many of these photos in a data set that helps work on computer vision.
43-year-old Mr. Geller wondered how his pictures could jump from place to place. They were then told that the photographs may have contributed to surveillance systems in the United States and other countries, and one of these systems was used to track China’s Uygur population.
“My eagerness turned to horror,” he said.
How the honeymoon photos helped build the surveillance system in China is, in some ways, a story of unintentional – or unexpected – consequences.
Several years ago, AI researchers from leading universities and tech companies started collecting digital photos from a wide variety of sources, including photo-sharing services, social networks, dating sites such as OkCupid and even college quids. Installed cameras were also included. He shared those photos with other organizations.
It was just ideal for researchers. They all needed data to feed into their new AI system, so they shared what they had. This was usually legal.
An example was Megaphase, a data set created in 2015 by professors at the University of Washington. They made it Without knowledge Or the consent of those whose pictures he turned into huge pools of his photographs. The professors posted it on the Internet so that others could download it.
According to a New York Times public record request, Megaface has been downloaded more than 6,000 times by companies and government agencies around the world. They included the US defense contractor Northrop Grumman; Q-Tel in Central Intelligence Agency’s investment arm; Bytdance, the parent company of the Chinese social media app TikTok; And Chinese surveillance company Megvi.
Researchers created MegaFace for use in an educational competition to inspire the development of facial recognition systems. It was not intended for commercial use. But only a few percent of those who downloaded Megaface publicly participated in the contest.
“We are not in a position to discuss third-party projects,” University of Washington spokesman Victor Balta said. “MegaFace has been decomposed, and MegaFace data is no longer being distributed.”
Some people who download data have deployed facial recognition systems. Megvi was blacklisted last year by the Department of Commerce after the Chinese government used its technology Monitor the Uighur population of the country.
The University of Washington took MegafaS offline in May, and other organizations have removed other data sets. But copies of these files can occur anywhere, and they are likely to feed new research.
Ms. O’Sullivan and Mr. Harvey spent years trying to create a device that could reveal how that data was being used. This was more difficult than anticipated.
They wanted to accept someone’s photo and, using facial recognition, immediately told the person how many times their face was included in their data set. But they worried that such equipment could be used in bad ways – by stalkers or companies and nation states.
“The potential for loss was very high,” said Ms. O’Sullivan, who is vice president of responsible AI with Arthur, a New York company that helps businesses manage the behavior of AI technologies.
Finally, they were forced to limit how people could discover the instrument and what the consequences were. The device, as it works today, is not as effective as they would like. But researchers worried that they could not uncover the breadth of the problem without making it worse.
Exposing.AI itself does not use facial recognition. It only pinpoints photos if you already have a way to point them online, with an Internet address. People can only search for photos that were posted to Flickr, and require a Flickr username, tag, or Internet address that can recognize those photos. (This provides appropriate security and privacy protection, the researchers said.)
Although this limits the utility of the device, it is still an eye opener. Make flickr images one Significant health of facial recognition data sets Which has been passed around the internet including megafaces.
It is not difficult to find photos that people have some personal relationship with. Just by searching through old emails for Flickr links, The Times landed photos that, according to Exposing.AI, were used in megafast and other facial recognition data sets.
Many belonged to Google’s well-known security researcher Parissa Tabriz. She did not respond to a request for comment.
He is particularly troubled by what Mr. Geller has discovered through the tool because he once believed that the free flow of information on the Internet was mostly a positive thing. He used Flickr because it empowered others to use his photos through a Creative Commons license.
“I’m living the result now,” he said.
Their hope – and Ms. O’Sullivan’s and Mr. Harvey’s hope – is that companies and governments will develop new norms, policies and laws that prevent large collections of personal data. He is making a documentary about the long, winding and sometimes disturbing way of his honeymoon photos to shed light on the problem.
Mr. Harvey is adamant that something needs to change. “We need to eliminate these as soon as possible – before they do more damage,” he said.