using more abstract features.) The system converts these details to a mathematical representation and compares them to data on other faces stored in a face recognition database. The data about a particular face is called a face template, and it can be compared to other templates on file. Facial recognition technology learns how to identify people by analyzing as many digital pictures as possible using neural networks, which are complex mathematical systems that require vast amounts of data to build pattern recognition. Face recognition tools are now frequently used in routine policing. Police compare mugshots of ar- restees to local, state, and federal face recognition databases. Law enforcement can query these mug- shot databases to identify people in photos taken from social media, traffic cameras, and closed circuit television surveillance cameras in stores, parks, and other places. There are systems to compare faces in real-time with "hot lists" of people suspected of il- legal activity. Face recognition has also been used in airports, border crossings, and events such as the Olympic games. The FBI spent more than a decade using such systems to compare driver's license and visa photos against the faces of suspected criminals. Facial recognition systems can make products safer and more secure. For example, face authentica- tion can ensure that only the right person gets ac- cess to sensitive information meant just for them. It can also be used for social good; there are nonprofits using facial recognition to combat trafficking of mi- nors. However these systems also have limitations that can do harm as well. Dozens of databases of people's faces are being compiled by companies and researchers, with many they do not distribute, according to research papers. But other companies and universities have widely shared their image troves with researchers, govern- ments, and private enterprises in Australia, China, India, Singapore, and Switzerland for training artifi- cial intelligence, according to academics, activists, and public papers. Startup Clearview AI created a powerful facial rec- ognition app that enables the user to take a picture of a person, upload it, and be able to view public photos of that person, along with links to where those pho- tos appeared. The system uses a database of more than three billion images that Clearview claims to have scraped from Facebook, YouTube, Venmo, and millions of other websites. Federal and state law enforcement officers have used the Clearview app to help solve shoplifting, identity theft, credit card fraud, murder, and child sexual exploitation cases. Companies and labs have gathered facial images for more than a decade, and image databases are an essential component of facial recognition technol- ogy. But people often have no idea that their faces are in them. And although names are typically not attached to the photos, individuals can be recognized because each face is unique to a person. There is no oversight of these facial recognition data repositories. Privacy advocates worry that facial recogni- tion systems are being misused. A database called Brainwash was created by Stanford University researchers in 2014. The researchers captured over 10,000 images using a camera located in San Francisco's Brainwash Caf (now closed). It is un- clear whether the patrons knew their images were being captured and used for research. The Stanford INTERACTIVE SESSION TECHNOLOGY Do You Know Who Is Using Your Face? Facial recognition is an artificial intelligence applica- of the images then being shared around the world. tion that can uniquely identify a person by analyz- The databases are pulled together with images from ing patterns based on the person's facial textures social networks, photo websites, dating services like and shape. Facial recognition systems can be used to OkCupid, and cameras placed in restaurants and on identify people in photos, video, or real-time. A fa- college quads. While there is no precise count of the cial recognition system uses biometrics to map facial data sets, privacy activists have pinpointed reposito- features from a photograph or video. It compares the ries that were built by Microsoft, Stanford University, information with a database of known faces to find a and others, with one holding over 10 million im- match. The face recognition system uses computer al- ages while another had more than two million. gorithms to highlight specific, distinctive details about Georgetown University has estimated that photos of a person's face, such as the distance between the eyes nearly half of all U.S. adults have been entered into or the shape of the chin. (Some algorithms explicitly at least one face recognition database. map the face, measuring the distances between the Tech giants like Facebook and Google are reputed eyes, nose, and mouth and so on. Others map the face to have amassed the largest facial data sets, which using more abstract features.) The system converts they do not distribute, according to research papers. these details to a mathematical representation and But other companies and universities have widely compares them to data on other faces stored in a face shared their image troves with researchers, govern- recognition database. The data about a particular face ments, and private enterprises in Australia, China, is called a face template, and it can be compared to India, Singapore, and Switzerland for training artifi- other templates on file. Facial recognition technology cial intelligence, according to academics, activists, learns how to identify people by analyzing as many and public papers. digital pictures as possible using neural networks, Startup Clearview AI created a powerful facial rec- which are complex mathematical systems that require ognition app that enables the user to take a picture of vast amounts of data to build pattern recognition. a person, upload it, and be able to view public photos Face recognition tools are now frequently used of that person, along with links to where those pho- in routine policing. Police compare mugshots of ar- tos appeared. The system uses a database of more restees to local, state, and federal face recognition than three billion images that Clearview claims to databases. Law enforcement can query these mug- have scraped from Facebook, YouTube, Venmo, and shot databases to identify people in photos taken millions of other websites. Federal and state law from social media, traffic cameras, and closed circuit enforcement officers have used the Clearview app 432 Part Three Key System Applications for the Digital Age researchers shared Brainwash with Chinese aca- demics associated with the National University of Defense Technology and Megvii, an AI company that provided surveillance technology for racial profiling of China's Uighur Muslim population. Brainwash was removed from its original website in mid-2019. Using eight cameras on campus to collect images, Duke University researchers gathered more than 2 million video frames with images of over 2,700 peo- ple. The database, called Duke MTMC, was reported to have been used to train Al systems in the United States, Japan, China, and elsewhere. The cameras were identified with signs, which gave a phone num- ber or email for people to opt out. Moreover, facial recognition systems are not entirely accurate. Face recognition systems have varying ability to identify people under challenging conditions such as poor lighting, low-quality image resolution, and suboptimal angle of view, which might occur if a photograph was taken from above looking down on an unknown person. Facial recognition software is poor at identifying African Americans and other ethnic minorities as well as women and young people. A 2012 study co- authored by the FBI reported that accuracy rates were lower for Afro-Americans than for other demograph- ics. Although the FBI claims that its facial recognition system can find the correct candidate in the top 50 profiles 85 percent of the time, that's only when the true candidate exists in its gallery. If the candidate is not in the gallery, the system may still come up with one or more potential matches, creating false positive results. Those identified could then be targeted as suspects for crimes they didn't commit. Face recognition becomes less accurate as the num- ber of people in the database increases. Many people around the world look alike. As the likelihood of simi- lar faces goes up, matching accuracy goes down. Sources: "Face Recognition," www.eff.org, accessed April 21, 2020; Kashmir Hill, "The Secretive Company that Might End Privacy as We Know It," New York Times, January 18, 2020; Cate Metz, "Facial Recognition Tech Is Growing Stronger, Thanks to Your Face," New York Times, July 13, 2019; www.ai.google.com, accessed April 21, 2020. CASE STUDY QUESTIONS 1. Explain the key technologies used in facial recog- nition systems. 2. What are the benefits of using facial recognition systems? How do they help organizations improve operations and decision making? What problems can they help solve? 3. Identify and describe the disadvantages of using facial recognition systems and facial databases