THE CHINESE government has made no secret of its extensive surveillance infrastructure. In the years since the Tiananmen Square protests, the Communist Party has used every high-tech tool at its disposal to record, monitor, and influence the behavior of 1.4 billion people. Advanced facial recognition software and artificial intelligence are just two of the latest initiatives touted in the state-run press. In some provinces, the Police Cloud program flags “abnormal” events in data aggregated from a dizzying range of sources: government and hospital logs, company records, closed-circuit TV footage, and internet and social media trawls. Skynet—its name an apparently unironic allusion to the rogue computer network in the Terminator franchise—tracks more than twenty million surveillance cameras; it can identify a person in seconds and is used to provide real-time security updates. Sharp Eyes, a nationwide system, connects private and commercial cameras to a centralized database. A recent operation involving drones disguised as flying birds—predominantly used to spy on the Uighur population in the Xinjiang autonomous region—reportedly goes by the code name Dove.1
Of course, as many journalists have noted, the country’s current setup is hardly as all-seeing as officials boast. But when the goal is social control, the illusion of omniscience is almost as effective as the real thing: you don’t need to watch everyone if you can get everyone to behave as though they’re being watched. In any case, according to the government’s own projections, reality will soon match the party’s proud rhetoric. Exact figures vary, but analysts estimate that as many as three hundred million cameras will be installed across the nation over the next two years, bringing the total number up to more than six hundred million, nearly one for every two people.
Many accounts of China’s quest to build the world’s most advanced surveillance system tend to depict its citizenry as passive subjects at the mercy of a repressive state. In the Western media, Big Brother analogies rightly, if tiresomely, abound.2 Little attention has been paid, however, to the quiet emergence of citizen-generated surveillance data. In the last few years, businesses, schools, and private individuals have begun streaming their closed-circuit camera footage onto websites for anyone to see. While it’s not unusual to store video footage on internet servers, China appears to lead the way when it comes to the amount of footage that is publicly accessible, perhaps because of the country’s looser norms around individual privacy and copyright. Some of the livestreaming appears to be consensual: online celebrities use it build their brand; some businesses broadcast as a form of advertising or additional security; and at least a few schools seem to do it as a courtesy to nervous parents. Other cases are less clear-cut. A retail employee on one popular feed, who was known for frequently picking her nose, seemed to be unaware that thousands of people were watching and talking about her in chatrooms.
Dragonfly Eyes (2017), the first feature film by the celebrated Chinese artist Xu Bing, grapples with these strange new conditions, which are transforming not just daily life but also the filmmaking process. To construct the movie, Xu programmed twenty computers to download some ten thousand of the millions of hours of footage available on popular surveillance websites. He and his assistants then spent eighteen months culling, reviewing, and editing the found footage into a loose narrative devised by the poet Zhai Yongming and the screenwriter Zhang Hanyi. “What I find most interesting about this project is that it seeks out a way of working and creating that matches this new kind of civilization,” Xu has explained. “There is no cameraperson on our team, but all of the 24-hour surveillance cameras across China and the world are working for us. This is a method of the present, but it also presages future artistic methods.”3
Xu’s shots are determined by the often canted-angle positioning of security cameras in badly lit settings. There are no actors, instead the film’s two protagonists—a provincial woman and her male stalker—are “played” by dozens of ordinary individuals, with dubbing and voiceover offering a sense of narrative continuity. The resulting story is a bizarre assemblage of footage from a strip of highway, a hospital dialysis unit, a dairy farm, and many other locations.
Strictly speaking, Xu’s film is not about top-down government surveillance—a choice that might have been dangerous for an artist who lives in Beijing and who, until recently, served as the vice president of the Central Academy of Fine Arts. But it is impossible to make a film about surveillance in China without evoking the Party’s oversight efforts. Long-standing programs to merge the government’s digital infrastructure with social media, residential cameras, and commercial surveillance systems—not to mention the fact that the police can already access footage from most cloud-based systems—have rendered any distinction between these categories meaningless.
Perhaps most disturbingly, Dragonfly Eyes reveals how the growing reach of the surveillance state is normalized and supplemented by countless individual acts of voluntary disclosure—a state of affairs that the legal scholar Bernard Harcourt has called the “expository society.”4 “Our understanding of surveillance for the most part is control, management and supervision,” Xu recently told the Wall Street Journal. “But these days, people’s relationship to surveillance is changing. In the past, it was the government using it. But now it’s expanded from the government to everyone.”5
These expository habits not only make us vulnerable to punishment but also produce new pleasures, identities, and desires. After completing a cut of the film, Xu and his crew tracked down many of the individuals who constituted his “cast.” (In yet another privacy violation, many cameras automatically GPS-tagged their footage.) All the subjects that he visited happily agreed to participate. Some people told Xu that they enjoyed watching the footage. The monks at a monastery told him that they share their rituals in order to teach viewers about the Buddhist way of life. One solitary shopkeeper said that his webcam provides him with his only meaningful connection to society.
DRAGONFLY EYES begins with a clever framing device that suggests the film was put together by a sentient computer program. In text-to-speech English, a robotic AI voiceover announces: “This is Dragonfly. Dragonfly has 28,000 eyes, blinking 40,000 times per second.” As it fast-forwards through myriad streams of footage, Dragonfly, hooked up to an object-recognition algorithm, attempts to tell us what it’s seeing, through the squares, lines, and tags it places over various images.
Just before the title credits appear, the glitchy videos slow down to show a woman, framed by a red box, walking by the side of a canal—marked on-screen as “river” and “water, transparent”—in what appears to be a cul-de-sac. The AI picks out a “white” house, “cement grey” stairs, “copper brown” cable, a “bricks red” wall, and a “cement grey” road. The woman suddenly falls off the road into the canal. For a while she stays afloat. Her head bobs in and out of the water, as the red lines of the object recognition algorithm scramble over it, suggesting the program is unsure what to make of what it’s seeing. Then she disappears.
For a few more minutes, the film proceeds by way of this desultory, associative logic, but eventually, as Dragonfly puts it, “the storyline swims into view.” The first half takes the form of an elliptical bildungsroman. We follow the composite “protagonist,” Qing Ting (which means “dragonfly” in Mandarin), as she leaves the Buddhist monastery where she’s been living to take up a series of low-paying jobs in the city. She first finds work at a highly mechanized dairy farm. There she is spotted (on the closed-circuit system that monitors its employees) by Ke Fan, her self-appointed male savior. A brief friendship develops, but they soon fall out. After she’s fired from the dairy farm, Qing Ting takes a job at a dry-cleaning business and, later, at a restaurant. One day on the way home from work, she passes by a plastic surgery clinic and gets offered a special discount. “I figured it out that in this society, you either need to change your mind or change your appearance,” she says. The transformation is an attempt to gain social mobility, but the audience may also read it as Xu’s acknowledgment that we don’t really know what she looks like in the first place.
If the film’s first act articulates a kind of surveillance-based social realism, the second veers toward absurdist parable. It is narrated by Ke Fan, who is convinced that surgery has turned Qing Ting into Xiao Xiao, a popular internet performer. After making fun of another celebrity’s looks, Xiao Xiao becomes the target of a toxic online hate campaign that pushes her offline and, it seems, to suicide. In a sort of “I am Spartacus” gesture, hundreds of women—and eventually Ke Fan—undergo plastic surgery to resemble Xiao Xiao. Paradoxically, it is this epidemic of virality that finally overwhelms the state’s surveillance apparatus. “That Xiao Xiao is turning up everywhere,” one police officer says. “If we can’t tell them apart, what chance do surveillance cameras have?”
Dragonfly Eyes is not a condemnation of the rise of ubiquitous surveillance. But Xu’s unusual approach ends up being more compelling, and more troubling, than that of a straightforward critique. Xu has developed a new form of narrative that captures the contradictions of surveillance technology, which at once pinpoints personal identities and submerges them in a mass of other averaged data points. Dragonfly Eyes similarly pulls us in two directions. On the one hand, our desire for meaning allows us to see the ever-changing faces of Qing Ting and Ke Fan as if they belonged to single, coherent characters. On the other, the plot makes a mockery of this human tendency. Its endless acts of substitution subvert the Truman Show-esque notion that every self-regarding person holds dear—namely, that one is the protagonist of one’s own life, with friends and family members playing supporting parts.
These changing technological conditions, Xu’s film suggests, revise nearly every aspect and assumption of conventional cinema. It makes little sense to call Dragonfly Eyes either a fictional narrative or a documentary, since its very premise of splicing the two together calls both of those terms into question. The film is better understood as an exploration of the new kinds of realism that our changing technological conditions make possible. Xu follows artists like Harun Farocki and Hito Steyerl in posing questions that weren’t conceivable a decade earlier. When everyone is their own content creator, what is the role of the director? In a world in which everything is already filmed, what kinds of artistic labor should be brought to bear on our surplus of footage? Or, as Xu asks in a title card at the start of the film, “When these seemingly random yet intricately connected clips are assembled, what’s the distance between the video fragments of real life and ‘reality?’”
AS A CONCEPTUAL artist, Xu has long held to the Zen teaching that “words are unreliable.” Much of his art is concerned with the political uses of language—and the fragility of the systems we use to represent it. These themes can be traced to his adolescence. Xu came of age during the Cultural Revolution, when Chinese characters were simplified and retooled as part of Mao’s “thought reform.”
His most famous work to date, Book from the Sky (1987-91), is a monumental installation made up of scrolls, wall panels, books, and wooden printing blocks. The paper surfaces are printed with ornate characters that at first glance seem to resemble traditional Chinese writing. But upon further inspection, Xu’s lexicon is revealed to be unreadable and meaningless, frustrating even the most educated readers.
Twenty-five years later, Xu devised an inverted Book from the Sky with his Book from the Ground (2012), a text that anyone could read. To construct the picture book, Xu collected examples of universal language—the ideograms, brand logos, pictograms, emoticons, and symbols that populate the airports, bathrooms, and trains of cosmopolitan cities. With these hyper-legible symbols, Xu tells the story of twenty-four hours in the life of an ordinary office worker.
For all its technological novelty, Dragonfly Eyes marks a continuation of these earlier projects, showing that moving images can be just as disturbingly unreliable as the written word. Like Book from the Ground, the film recycles a found visual language to tell a cheeky story. At the same time, like Book from the Sky, the movie invites viewers to impose their own meaning on found footage that might seem arbitrary otherwise. But in Dragonfly Eyes, Xu pushes his language games even further—playing not just with known characters and symbols but with real people’s lives.
Just as his early work evoked the arbitrary dictums of the Cultural Revolution, Xu’s artful attempt to wring order from chaotic streams of footage calls to mind the state’s current directorial powers. After all, Xu’s film itself operates as a surveillance program of sorts, in that it makes meaning from massive troves of data. In the future one could even imagine Xu eliminating the need for a voiceover, choosing instead to use facial recognition and GPS monitoring, as Police Cloud does, to track specific individuals throughout their days. Xu’s film may be harmless fiction, driven by a playful sensibility, but it serves as a reminder of what might happen should the government use its vast troves of footage to spin more pernicious narratives.
In 2015 police shot a peasant man dead at a railway station in Heilongjiang, a province in the far sub-Siberian north. As Chaohua Wang reported in the London Review of Books, eyewitnesses quickly posted photos of the dead man online. Local police, in turn, “demanded that all photos and video clips of the killing be deleted,” but they were moments too late. The images that the public had seen of the murder led to outrage online and even physical protests. Government officials in Beijing soon intervened. China Central Television ran a story on the case in which video footage from several surveillance cameras at the station was edited together in an effort to portray the dead man as a drunk who had grabbed a police baton. Lawyers, meanwhile, were denied access to the original, unedited surveillance footage. The government’s goal was not to get the public to believe the official story so much as to get them to distrust the evidence gathered by civilians. “Doubt,” Wang wrote, “had been sown about the truth of the popular perception of the case.”6 Now that the state’s dragonfly eyes are always watching, there will only be more footage, more evidence, more editing, more uncertainty.
Aren’t we all trying to affect a projection of an authentic self? These works suggest that there is no such thing. Read more
In an old-factory-turned-hi-tech-art-space in Chelsea, the data paintings are almost absurdly moving—flickering, colorful nostalgia not for a New York that no longer exists, but one that never was. Read more
If the future of creativity is one where AI tools determine the types of art that gets made, then the future looks bleak. Read more