Sidney Perkowitz September 28 2020

Facing up to facial recognition

Films

Contemplating your face in the mirror as you brush your teeth in the morning, do you consider its role as you make it through the day? Everyone you encounter, from your spouse to strangers in the street or on Zoom, sees how you present yourself and could make a good guess about your mood. Recognizing faces and the emotions they show is a highly evolved human ability. Computers can also recognize a face, in order to unlock a phone or identify criminals and suspects as seen by surveillance cameras. But algorithmic facial identification by police is now under fire because it is less accurate for Black faces than white ones, causing biased false arrests or worse. Several LaboCine films treat facial recognition, both the algorithmic and human kinds.

Slaughterbots begins with a man dressed like a tech company executive and standing on stage before an audience, but this is no TED talk. He is selling a new law enforcement and military tool, a tiny palm-sized autonomous drone. It is a smart weapon that decides when it has found the right target to kill, using an onboard camera and facial recognition technology to identify a pre-programmed person. The presenter demonstrates how the drone homes in on a dummy standing on stage and delivers death with a small shaped charge that drills right through the temple into the brain. The presentation continues with a video showing real people being executed by the drones; all “bad guys” as the speaker reassures the audience.

These drones can be released in unstoppable swarms. The film moves into pseudo documentary mode when it shows imaginary TV news about drone swarms killing 11 selected U. S. Senators in the Capitol Building, all from the same party, and killing thousands of university student activists who were targeted through social media. When we return to the presenter, he is ending his demonstration by saying to a wildly cheering audience, “Smart weapons consume data. When you can find your enemy using data…you can target an evil ideology [pointing to his temple] right where it starts.” A real-life postscript follows as Stuart Russell, an AI researcher at UC – Berkeley, explains that the film is a warning against the dangers of allowing “machines to choose to kill humans” – a warning echoed by many in the AI community.

Slaughterbots

What Is My Face? is a welcome counterpoint to what is projected in Slaughterbots because it deals with the benign and fascinating topic of human face recognition. It begins with a request: “Please describe your face,” followed by the faces of different people who struggle to answer that seemingly simple query.

Two of the faces belong to Sofía Landi, a neuroscientist at The Rockefeller University in New York, and Mark Slutsky, a Montreal-based writer and filmmaker. Together they created this film to show the complexities of faces from different viewpoints. Slutsky’s view says much about the human impact of faces: “The essential element of a film is a close-up of a person’s face,” he says. “It’s how you sort of create emotion in movies, visually it’s usually through people’s faces.” Landi studies face recognition, especially “how our brains allow us to recognize people we know... Humans and other primates have areas in their brains that respond selectively to faces.” In 2017, Landi reported on facial recognition in macaque monkeys. Using functional magnetic resonance imaging (fMRI), which shows brain activity nearly in real time, she found that familiar faces excite two previously unknown brain areas that extend the brain’s known face-processing system.

Landi and Slutsky go on to tell us more about facial recognition, illustrated with images of Landi’s research and faces in crowds. Slutsky underlines the power of faces by pointing out that we see them in nonhuman objects like the Man in the Moon, and the remarkable fact that even with radically scrambled features as in a Picasso painting, we know we are seeing a face. Landi explains that it’s hard to describe a face because our brains take a holistic view, and describes the steps in facial recognition: finding the face in a scene, identifying the person behind the face, and reading the emotions the face expresses. But, she adds, “there is still so much we don’t understand.” At the film’s end, a street artist caricatures Landis and Slutsky together and sums it all up when he says “A face is your identity and your signature. Everything [is in] your face.”

What Is My Face?

In Vivid Detail shows what happens when the “everything” a face projects is not received. We first see Justin (John Ventimiglia) on lunch break, trying to assemble a picture puzzle of a smiling emoji face. At the architectural firm where he works, he meets consultant Leslie (Piper Perabo), a woman with an appealing face. After some office interactions, she asks if he wants to discuss his project over lunch, but he begs off. He didn’t understand that she was showing interest in him until a co-worker says, “Justin, she asked you to lunch.” Later he awkwardly asks Leslie to dinner. It goes well and they share a kiss afterwards, but two puzzling incidents occur. He mistakes one blonde waitress for another, and when two of Justin’s friends appear, it’s obvious to Leslie and to us that they are identical twins, but not to Justin.

The next day, Leslie is hurt when Jason seems to ignore her at work. He finally explains that he sees faces like stick figures that lack unique features. He cannot recognize people when, for instance, they change hair style. He has had this neurological condition, prosopagnosia, ever since he injured a particular part of his brain at age five. It’s like being colorblind, he says, where “you can see colors but you can’t tell them apart.” Leslie seems to accept this and they are playful about his inability to see faces. But when he has to admit that he can’t even tell when she is smiling, they both get angry and frustrated and Leslie leaves.

Later, as Jason walks alone, he sees a street artist carefully sketching a child’s face and gets an idea. He asks Leslie to pose for him. She agrees but looks dubious and asks “What are you doing?” He says “I want to be able to see you – right now,” as he carefully draws her face, detail by detail and square by square, on a cross-hatched surface. And when he asks her to smile, she responds with a gorgeous and radiant smile.

In Vivid Detail

What Is My Face? shows the complexities of facial recognition and that it will take more research to fully understand it. This makes it all the more remarkable that our eyes and brain carry out this intricate task without conscious effort, making visual recognition a vital part of human interaction. In Vivid Detail indicates just how vital by showing that those with face blindness or prosopagnosia, about 2 percent of the population, lose the essential social and emotional benefits of reading faces. Algorithmic facial recognition as seen in the science fiction world of Slaughterbots, however, and in our real world of policing, seems to offer more costs than benefits. We need to think less about extending this technology and more about the ethics of using it.

About the author

Sidney Perkowitz, Candler Professor of Physics Emeritus at Emory University, writes often about science in film. His latest books are Physics A Very Short Introduction and Real Scientists Don’t Wear Ties.

 

Download Labocine's iOS App