Citing National Security, Researchers Race to Foil ‘Deepfake’ Videos
As next year’s presidential election inches closer, researchers are working to develop systems to more quickly and accurately detect “deepfakes” — digitally-altered and highly-realistic videos capable of making politicians or other high-profile individuals appear to be saying whatever the creator dreams up.
Such videos employ a host of artificial intelligence (AI) tools to turn people into virtual puppets, from face-swapping and lip-syncing to generating speech using a target’s real voice. Just this week, artists released a video that appears to show Mark Zuckerberg delivering a sinister message about the power afforded to him by Facebook. And in April of last year, filmmaker Jordan Peele provided the voice for a public service announcement that appeared to be coming directly from former U.S. President Barack Obama. While both videos were presented transparently with an aim to raise awareness, a paper published by researchers in California on Wednesday warns that it could be just a matter of time before the technology is used for more nefarious purposes, posing “a significant threat to our democracy, national security, and society.
“With relatively modest amounts of data and computing power,” the authors wrote, “the average person can, for example, create a video of a world leader confessing to illegal activity leading to a constitutional crisis, a military leader saying something racially insensitive leading to civil unrest in an area of military activity, or a corporate titan claiming that their profits are weak leading to global stock manipulation.”
In an effort to guard against this, the research team, with funding from Google, Microsoft, and the U.S. Department of Defense, developed a system to break down and encode an individual’s facial expressions and movements in each frame in a video. These so-called “facial action units,” include the specific ways individuals raise their eyebrows, parts their lips, and move their jaws. Coupling these with measurements including the distance between the corners of a person’s mouth, the technology was able to accurately detect deepfakes created by swapping the faces of Elizabeth Warren, Hillary Clinton, and President Trump, among others, onto their Saturday Night Live impersonators.
Still, even promising detection methods are likely to be useful only for so long. “This is a cat-and-mouse game,” David Doermann, a computer scientist and media forensics expert at the University at Buffalo said at a House intelligence committee meeting on Thursday. “As things get better for being able to deceive visually, [manipulators are] going to move on to covering up their trace evidence.”
Hany Farid, a computer science professor at UC Berkeley and a member of the research team, echoed that sentiment: “We are outgunned,” he told The Washington Post. “The number of people working on the video-synthesis side, as opposed to the detector side, is 100 to 1.”
Aside from detection, there’s also the question of what social media platforms should do when a fake is detected. In late May, an edited video began circulating that purported to show House Speaker Nancy Pelosi slurring her words during a speech. More of a “cheapfake” than a “deepfake,” the video still garnered millions of views and was even shared by President Trump. And while YouTube chose to remove it, the video remains available on Facebook and Twitter.
Also in the news:
• The movement to end all-male scientific panels, or manels, has a new high-profile advocate. Dr. Francis Collins, director of the National Institutes of Health, announced this week that he would no longer speak at conferences that do not show a strong commitment to including women and members of other underrepresented groups in speaking roles. The announcement was widely applauded by the science community. “People really want him at a conference,” Princeton neuroscientist Yael Niv told The New York Times. “So if he says, ‘I’m not coming to your conference to give the keynote speech because I don’t see adequate representation,’ that is huge.” In a statement released Wednesday, Collins cited a recent report by the National Academy of Sciences, Engineering, and Medicine, which urged scientific leaders to help “combat the cultural forces that tolerate gender harassment and limit the advancement of women.” Collins wrote that “starting now, when I consider speaking invitations … if that attention to inclusiveness is not evident in the agenda, I will decline to take part.” Within hours, other prominent men in science signaled they would do the same. (The New York Times)
• Americans are now consuming tens of thousands of microplastic particles every year, according to a new analysis published in the journal Environmental Science and Technology. The study, led by a team of Canadian scientists, got widespread attention in part because of the wealth of data analyzed – the researchers used more than 400 data points in a review of some 26 previously published studies. By their analysis, Americans swallow between 39,000 and 52,000 particles a year – the variation is explained by age, gender, and diet. Fish, for instance, tend to be higher in plastics because of the amount that is dumped and scattered into the ocean. The scientists noted that the exposure is even higher if one considers the plastics that we inhale. And it jumps up again for people who drink water, sodas, and other beverages from plastic bottles. The study suggested that a person drinking only bottled water would consume at least an additional 90,000 particles of microplastics every year. Scientists are still trying to understand the health effects of plasticized diet but early research suggests possible impacts on both the immune system and the gut microbiome. Study author Kieran Cox emphasized that the primary point of a study was a reminder that humans are not somehow buffered from the spread of plastics across the planet’s environment. (National Geographic)
• Psilocybin, the naturally-occurring compound that gives magic mushrooms their magic, has shown promise as a treatment for severe depression in a small but growing number of studies. Among other trials sponsored by universities and medical organizations, this January, Imperial College London began a new experiment to dose 60 patients with psilocybin and compare their progress to a group receiving escitalopram, a common antidepressant. Robin Carhart-Harris, a lead researcher on the Imperial College trial, says he believes the psilocybin’s effectiveness is due in part to the fact that its psychedelic properties allow patients to see their lives more clearly, perhaps allowing them to confront the sources of their mental illness. Long criminalized in many countries for its mind-altering effects and its association with counterculture, the neurological properties of psychedelics have yet to be fully explored, but scans of patients’ brains after being treated with psilocybin showed reduced blood flow to the amygdala, which is often overactive in mental illnesses like depression. Recently, Denver, Colorado and Oakland, California moved to decriminalize magic mushrooms, which researchers see as a step towards opening up new investigations on psilocybin’s potential as a therapeutic. (The Guardian)
• Japan will resume commercial whaling activities next month, withdrawing from a 1986 international accord that established a global moratorium on whale hunting. While the International Whaling Commission (IWC) allowed Japan to continue killing whales under the auspices of “scientific research,” the commercial whaling ban has been widely regarded as a successful and necessary conservation measure. Japanese whale meat consumption decreased from 200,000 tons in the 1960s to just 3,000 tons in 2016. Yoshihide Suga, Japan’s chief cabinet secretary, criticized the IWC for being overly focused on conservation instead of developing a sustainable whaling industry, and issued a statement claiming that commercial whaling has historical and cultural significance for Japanese communities in addition to providing a desired source of protein. When Japan first announced its intention to withdraw from the IWC last year, conservationists strongly condemned the move as unnecessary and harmful to whale populations that are already struggling. Greenpeace Japan executive director Sam Annesley responded with a call for a reversal of Japan’s withdraw. “The world’s oceans face multiple threats such as acidification and plastic pollution, in addition to overfishing,” he said in a December 2018 statement. “As a country surrounded by oceans where people’s lives have been heavily reliant on marine resources, it is essential for Japan to work towards healthy oceans. Japan’s government has so far failed to resolve these problems.” (Quartz)
• And finally: In news likely to delight anthropologists, archaeologists, and jam-band enthusiasts in equal measure, scientists announced this week that they had uncovered the first-ever direct evidence that people have been using cannabis to get high at least as far back as Bronze Age China — or 2,500 B.C. That assertion, published this week in the journal Science Advances, is tied to the discovery of several wooden stone-filled bowls excavated from burial sites at the Jirzankal Cemetery in western China. When the stones were analyzed, National Geographic reported, the researchers uncovered “molecular remnants of tetrahydrocannabinol, or THC — the chemical responsible for cannabis’ psychoactive effects.” From this they theorized that ritualistic, funerary inhalation of the drug was almost certainly the intent. “In the Jirzankal Cemetery,” Wired magazine reported, “… someone was blazing.” That such activity was already established prior to the common era had, until now, only been indirectly suggested by a single passage from the 5th century Greek historian Herodotus, who wrote of contemporaneous Scythians burning cannabis on hot stones and “howl[ing] in their joy at the vapor-bath.” (Science Advances and National Geographic)