“Deep Fake” Video Technology Is Advancing Faster Than Our Policies Can Keep Up
This article is part of the magazine, "The Future of Science In America: The Election Issue," co-published by LeapsMag, the Aspen Institute Science & Society Program, and GOOD.
Alethea.ai sports a grid of faces smiling, blinking and looking about. Some are beautiful, some are oddly familiar, but all share one thing in common—they are fake.
Alethea creates "synthetic media"— including digital faces customers can license saying anything they choose with any voice they choose. Companies can hire these photorealistic avatars to appear in explainer videos, advertisements, multimedia projects or any other applications they might dream up without running auditions or paying talent agents or actor fees. Licenses begin at a mere $99. Companies may also license digital avatars of real celebrities or hire mashups created from real celebrities including "Don Exotic" (a mashup of Donald Trump and Joe Exotic) or "Baby Obama" (a large-eared toddler that looks remarkably similar to a former U.S. President).
Naturally, in the midst of the COVID pandemic, the appeal is understandable. Rather than flying to a remote location to film a beer commercial, an actor can simply license their avatar to do the work for them. The question is—where and when this tech will cross the line between legitimately licensed and authorized synthetic media to deep fakes—synthetic videos designed to deceive the public for financial and political gain.
Deep fakes are not new. From written quotes that are manipulated and taken out of context to audio quotes that are spliced together to mean something other than originally intended, misrepresentation has been around for centuries. What is new is the technology that allows this sort of seamless and sophisticated deception to be brought to the world of video.
"At one point, video content was considered more reliable, and had a higher threshold of trust," said Alethea CEO and co-founder, Arif Khan. "We think video is harder to fake and we aren't yet as sensitive to detecting those fakes. But the technology is definitely there."
"In the future, each of us will only trust about 15 people and that's it," said Phil Lelyveld, who serves as Immersive Media Program Lead at the Entertainment Technology Center at the University of Southern California. "It's already very difficult to tell true footage from fake. In the future, I expect this will only become more difficult."
How do we know what's true in a world where original videos created with avatars of celebrities and politicians can be manipulated to say virtually anything?
As the U.S. 2020 Presidential Election nears, the potential moral and ethical implications of this technology are startling. A number of cases of truth tampering have recently been widely publicized. On August 5, President Donald Trump's campaign released an ad featuring several photos of Joe Biden that were altered to make it seem like was hiding all alone in his basement. In one photo, at least ten people who had been sitting with Biden in the original shot were cut out. In other photos, Biden's image was removed from a nature preserve and praying in church to make it appear Biden was in that same basement. Recently several videos of Speaker of the House Nancy Pelosi were slowed down by 75 percent to make her sound as if her speech was slurred.
During a campaign event in Florida on September 15 of this year, former Vice President Joe Biden was introduced by Puerto Rican singer-songwriter Luis Fonsi. After he was introduced, Biden paid tribute to the singer-songwriter—he held up his cell phone and played the hit song "Despecito". Shortly afterward, a doctored version of this video appeared on self-described parody site the United Spot replacing the Despicito with N.W.A.'s "F—- Tha Police". By September 16, Donald Trump retweeted the video, twice—first with the line "What is this all about" and second with the line "China is drooling. They can't believe this!" Twitter was quick to mark the video in these tweets as manipulated media.
Twitter had previously addressed several of Donald Trump's tweets—flagging a video shared in June as manipulated media and removing altogether a video shared by Trump in July showing a group promoting the hydroxychloroquine as an effective cure for COVID-19. Many of these manipulated videos are ultimately flagged or taken down, but not before they are seen and shared by millions of online viewers.
These faked videos were exposed rather quickly, as they could be compared with the original, publicly available source material. But what happens when there is no original source material? How do we know what's true in a world where original videos created with avatars of celebrities and politicians can be manipulated to say virtually anything?
"This type of fake media is a profound threat to our democracy," said Reid Blackman, the CEO of VIRTUE--an ethics consultancy for AI leaders. "Democracy depends on well-informed citizens. When citizens can't or won't discern between real and fake news, the implications are huge."
In light of the importance of reliable information in the political system, there's a clear and present need to verify that the images and news we consume is authentic. So how can anyone ever know that the content they are viewing is real?
"This will not be a simple technological solution," said Blackman. "There is no 'truth' button to push to verify authenticity. There's plenty of blame and condemnation to go around. Purveyors of information have a responsibility to vet the reliability of their sources. And consumers also have a responsibility to vet their sources."
Yet the process of verifying sources has never been more challenging. More and more citizens are choosing to live in a "media bubble"—gathering and sharing news only from and with people who share their political leanings and opinions. At one time, United States broadcasters were bound by the Fairness Doctrine—requiring them to present controversial issues important to the public in a way that the FCC deemed honest, equitable and balanced. The repeal of this doctrine in 1987 paved the way for new forms of cable news channels such as Fox News and MSNBC that appealed to viewers with a particular point of view. The Internet has only exacerbated these tendencies. Social media algorithms are designed to keep people clicking within their comfort zones by presenting members with only the thoughts and opinions they want to hear.
"I sometimes laugh when I hear people tell me they can back a particular opinion they hold with research," said Blackman. "Having conducted a fair bit of true scientific research, I am aware that clicking on one article on the Internet hardly qualifies. But a surprising number of people believe that finding any source online that states the fact they choose to believe is the same as proving it true."
Back to the fundamental challenge: How do we as a society root out what's false online? Lelyveld suggests that it will begin by verifying things that are known to be true rather than trying to call out everything that is fake. "The EU called me in to talk about how to deal with fake news coming out of Russia," said Lelyveld. "I told them Hollywood has spent 100 years developing special effects technology to make things that are wholly fictional indistinguishable from the truth. I told them that you'll never chase down every source of fake news. You're better off focusing on what can be proved true."
Arif Khan agrees. "There are probably 100 accounts attributed to Elon Musk on Twitter, but only one has the blue checkmark," said Khan. "That means Twitter has verified that an account of public interest is real. That's what we're trying to do with our platform. Allow celebrities to verify that specific videos were licensed and authorized directly by them."
Alethea will use another key technology called blockchain to mark all authentic authorized videos with celebrity avatars. Blockchain uses a distributed ledger technology to make sure that no undetected changes have been made to the content. Think of the difference between editing a document in a traditional word processing program and editing in a distributed online editing system like Google Docs. In a traditional word processing program, you can edit and copy a document without revealing any changes. In a shared editing system like Google Docs, every person who shares the document can see a record of every edit, addition and copy made of any portion of the document. In a similar way, blockchain helps Alethea ensure that approved videos have not been copied or altered inappropriately.
While AI companies like Alethea are moving to ensure that avatars based on real individuals aren't wrongly identified, the situation becomes a bit murkier when it comes to the question of representing groups, races, creeds, and other forms of identity. Alethea is rightly proud that the completely artificial avatars visually represent a variety of ages, races and sexes. However, companies could conceivably license an avatar to represent a marginalized group without actually hiring a person within that group to decide what the avatar will do or say.
"I don't know if I would call this tokenism, as that is difficult to identify without understanding the hiring company's intent," said Blackman. "Where this becomes deeply troubling is when avatars are used to represent a marginalized group without clearly pointing out the actor is an avatar. It's one thing for an African American woman avatar to say, 'I like ice cream.' It's entirely different thing for an African American woman avatar to say she supports a particular political candidate. In the second case, the avatar is being used as social proof that real people of a certain type back a certain political idea. And there the deception is far more problematic."
"It always comes down to unintended consequences of technology," said Lelyveld. "Technology is neutral—it's only the implementation that has the power to be good or bad. Without a thoughtful approach to the cultural, moral and political implications of technology, it often drifts towards the bad. We need to make a conscious decision as we release new technology to ensure it moves towards the good."
When presented with the idea that his avatars might be used to misrepresent marginalized groups, Khan was thoughtful. "Yes, I can see that is an unintended consequence of our technology. We would like to encourage people to license the avatars of real people, who would have final approval over what their avatars say or do. As to what people do with our completely artificial avatars, we will have to consider that moving forward."
Lelyveld frankly sees the ability for advertisers to create avatars that are our assistants or even our friends as a greater moral concern. "Once our digital assistant or avatar becomes an integral part of our life—even a friend as it were, what's to stop marketers from having those digital friends make suggestions about what drink we buy, which shirt we wear or even which candidate we elect? The possibilities for bad actors to reach us through our digital circle is mind-boggling."
Ultimately, Blackman suggests, we as a society will need to make decisions about what matters to us. "We will need to build policies and write laws—tackling the biggest problems like political deep fakes first. And then we have to figure out how to make the penalties stiff enough to matter. Fining a multibillion-dollar company a few million for a major offense isn't likely to move the needle. The punishment will need to fit the crime."
Until then, media consumers will need to do their own due diligence—to do the difficult work of uncovering the often messy and deeply uncomfortable news that's the truth.
[Editor's Note: To read other articles in this special magazine issue, visit the beautifully designed e-reader version.]
Scientists Are Working to Decipher the Puzzle of ‘Broken Heart Syndrome’
Elaine Kamil had just returned home after a few days of business meetings in 2013 when she started having chest pains. At first Kamil, then 66, wasn't worried—she had had some chest pain before and recently went to a cardiologist to do a stress test, which was normal.
"I can't be having a heart attack because I just got checked," she thought, attributing the discomfort to stress and high demands of her job. A pediatric nephrologist at Cedars-Sinai Hospital in Los Angeles, she takes care of critically ill children who are on dialysis or are kidney transplant patients. Supporting families through difficult times and answering calls at odd hours is part of her daily routine, and often leaves her exhausted.
She figured the pain would go away. But instead, it intensified that night. Kamil's husband drove her to the Cedars-Sinai hospital, where she was admitted to the coronary care unit. It turned out she wasn't having a heart attack after all. Instead, she was diagnosed with a much less common but nonetheless dangerous heart condition called takotsubo syndrome, or broken heart syndrome.
A heart attack happens when blood flow to the heart is obstructed—such as when an artery is blocked—causing heart muscle tissue to die. In takotsubo syndrome, the blood flow isn't blocked, but the heart doesn't pump it properly. The heart changes its shape and starts to resemble a Japanese fishing device called tako-tsubo, a clay pot with a wider body and narrower mouth, used to catch octopus.
"The heart muscle is stunned and doesn't function properly anywhere from three days to three weeks," explains Noel Bairey Merz, the cardiologist at Cedar Sinai who Kamil went to see after she was discharged.
"The heart muscle is stunned and doesn't function properly anywhere from three days to three weeks."
But even though the heart isn't permanently damaged, mortality rates due to takotsubo syndrome are comparable to those of a heart attack, Merz notes—about 4-5 percent of patients die from the attack, and 20 percent within the next five years. "It's as bad as a heart attack," Merz says—only it's much less known, even to doctors. The condition affects only about 1 percent of people, and there are around 15,000 new cases annually. It's diagnosed using a cardiac ventriculogram, an imaging test that allows doctors to see how the heart pumps blood.
Scientists don't fully understand what causes Takotsubo syndrome, but it usually occurs after extreme emotional or physical stress. Doctors think it's triggered by a so-called catecholamine storm, a phenomenon in which the body releases too much catecholamines—hormones involved in the fight-or-flight response. Evolutionarily, when early humans lived in savannas or forests and had to either fight off predators or flee from them, these hormones gave our ancestors the needed strength and stamina to take either action. Released by nerve endings and by the adrenal glands that sit on top of the kidneys, these hormones still flood our bodies in moments of stress, but an overabundance of them could sometimes be damaging.
Elaine Kamil
A study by scientists at Harvard Medical School linked increased risk of takotsubo to higher activity in the amygdala, a brain region responsible for emotions that's involved in responses to stress. The scientists believe that chronic stress makes people more susceptible to the syndrome. Notably, one small study suggested that the number of Takotsubo cases increased during the COVID-19 pandemic.
There are no specific drugs to treat takotsubo, so doctors rely on supportive therapies, which include medications typically used for high blood pressure and heart failure. In most cases, the heart returns to its normal shape within a few weeks. "It's a spontaneous recovery—the catecholamine storm is resolved, the injury trigger is removed and the heart heals itself because our bodies have an amazing healing capacity," Merz says. It also helps that tissues remain intact. 'The heart cells don't die, they just aren't functioning properly for some time."
That's the good news. The bad news is that takotsubo is likely to strike again—in 5-20 percent of patients the condition comes back, sometimes more severe than before.
That's exactly what happened to Kamil. After getting her diagnosis in 2013, she realized that she actually had a previous takotsubo episode. In 2010, she experienced similar symptoms after her son died. "The night after he died, I was having severe chest pain at night, but I was too overwhelmed with grief to do anything about it," she recalls. After a while, the pain subsided and didn't return until three years later.
For weeks after her second attack, she felt exhausted, listless and anxious. "You lose confidence in your body," she says. "You have these little twinges on your chest, or if you start having arrhythmia, and you wonder if this is another episode coming up. It's really unnerving because you don't know how to read these cues." And that's very typical, Merz says. Even when the heart muscle appears to recover, patients don't return to normal right away. They have shortens of breath, they can't exercise, and they stay anxious and worried for a while.
Women over the age of 50 are diagnosed with takotsubo more often than other demographics. However, it happens in men too, although it typically strikes after physical stress, such as a triathlon or an exhausting day of cycling. Young people can also get takotsubo. Older patients are hospitalized more often, but younger people tend to have more severe complications. It could be because an older person may go for a jog while younger one may run a marathon, which would take a stronger toll on the body of a person who's predisposed to the condition.
Notably, the emotional stressors don't always have to be negative—the heart muscle can get out of shape from good emotions, too. "There have been case reports of takotsubo at weddings," Merz says. Moreover, one out of three or four takotsubo patients experience no apparent stress, she adds. "So it could be that it's not so much the catecholamine storm itself, but the body's reaction to it—the physiological reaction deeply embedded into out physiology," she explains.
Merz and her team are working to understand what makes people predisposed to takotsubo. They think a person's genetics play a role, but they haven't yet pinpointed genes that seem to be responsible. Genes code for proteins, which affect how the body metabolizes various compounds, which, in turn, affect the body's response to stress. Pinning down the protein involved in takotsubo susceptibility would allow doctors to develop screening tests and identify those prone to severe repeating attacks. It will also help develop medications that can either prevent it or treat it better than just waiting for the body to heal itself.
Researchers at the Imperial College London found that elevated levels of certain types of microRNAs—molecules involved in protein production—increase the chances of developing takotsubo.
In one study, researchers tried treating takotsubo in mice with a drug called suberanilohydroxamic acid, or SAHA, typically used for cancer treatment. The drug improved cardiac health and reversed the broken heart in rodents. It remains to be seen if the drug would have a similar effect on humans. But identifying a drug that shows promise is progress, Merz says. "I'm glad that there's research in this area."
This article was originally published by Leaps.org on July 28, 2021.
Lina Zeldovich has written about science, medicine and technology for Popular Science, Smithsonian, National Geographic, Scientific American, Reader’s Digest, the New York Times and other major national and international publications. A Columbia J-School alumna, she has won several awards for her stories, including the ASJA Crisis Coverage Award for Covid reporting, and has been a contributing editor at Nautilus Magazine. In 2021, Zeldovich released her first book, The Other Dark Matter, published by the University of Chicago Press, about the science and business of turning waste into wealth and health. You can find her on http://linazeldovich.com/ and @linazeldovich.
Did Anton the AI find a new treatment for a deadly cancer?
Bile duct cancer is a rare and aggressive form of cancer that is often difficult to diagnose. Patients with advanced forms of the disease have an average life expectancy of less than two years.
Many patients who get cancer in their bile ducts – the tubes that carry digestive fluid from the liver to the small intestine – have mutations in the protein FGFR2, which leads cells to grow uncontrollably. One treatment option is chemotherapy, but it’s toxic to both cancer cells and healthy cells, failing to distinguish between the two. Increasingly, cancer researchers are focusing on biomarker directed therapy, or making drugs that target a particular molecule that causes the disease – FGFR2, in the case of bile duct cancer.
A problem is that in targeting FGFR2, these drugs inadvertently inhibit the FGFR1 protein, which looks almost identical. This causes elevated phosphate levels, which is a sign of kidney damage, so doses are often limited to prevent complications.
In recent years, though, a company called Relay has taken a unique approach to picking out FGFR2, using a powerful supercomputer to simulate how proteins move and change shape. The team, leveraging this AI capability, discovered that FGFR2 and FGFR1 move differently, which enabled them to create a more precise drug.
Preliminary studies have shown robust activity of this drug, called RLY-4008, in FGFR2 altered tumors, especially in bile duct cancer. The drug did not inhibit FGFR1 or cause significant side effects. “RLY-4008 is a prime example of a precision oncology therapeutic with its highly selective and potent targeting of FGFR2 genetic alterations and resistance mutations,” says Lipika Goyal, assistant professor of medicine at Harvard Medical School. She is a principal investigator of Relay’s phase 1-2 clinical trial.
Boosts from AI and a billionaire
Traditional drug design has been very much a case of trial and error, as scientists investigate many molecules to see which ones bind to the intended target and bind less to other targets.
“It’s being done almost blindly, without really being guided by structure, so it fails very often,” says Olivier Elemento, associate director of the Institute for Computational Biomedicine at Cornell. “The issue is that they are not sampling enough molecules to cover some of the chemical space that would be specific to the target of interest and not specific to others.”
Relay’s unique hardware and software allow simulations that could never be achieved through traditional experiments, Elemento says.
Some scientists have tried to use X-rays of crystallized proteins to look at the structure of proteins and design better drugs. But they have failed to account for an important factor: proteins are moving and constantly folding into different shapes.
David Shaw, a hedge fund billionaire, wanted to help improve drug discovery and understood that a key obstacle was that computer models of molecular dynamics were limited; they simulated motion for less than 10 millionths of a second.
In 2001, Shaw set up his own research facility, D.E. Shaw Research, to create a supercomputer that would be specifically designed to simulate protein motion. Seven years later, he succeeded in firing up a supercomputer that can now conduct high speed simulations roughly 100 times faster than others. Called Anton, it has special computer chips to enable this speed, and its software is powered by AI to conduct many simulations.
After creating the supercomputer, Shaw teamed up with leading scientists who were interested in molecular motion, and they founded Relay Therapeutics.
Elemento believes that Relay’s approach is highly beneficial in designing a better drug for bile duct cancer. “Relay Therapeutics has a cutting-edge approach for molecular dynamics that I don’t believe any other companies have, at least not as advanced.” Relay’s unique hardware and software allow simulations that could never be achieved through traditional experiments, Elemento says.
How it works
Relay used both experimental and computational approaches to design RLY-4008. The team started out by taking X-rays of crystallized versions of both their intended target, FGFR2, and the almost identical FGFR1. This enabled them to get a 3D snapshot of each of their structures. They then fed the X-rays into the Anton supercomputer to simulate how the proteins were likely to move.
Anton’s simulations showed that the FGFR1 protein had a flap that moved more frequently than FGFR2. Based on this distinct motion, the team tried to design a compound that would recognize this flap shifting around and bind to FGFR2 while steering away from its more active lookalike.
For that, they went back Anton, using the supercomputer to simulate the behavior of thousands of potential molecules for over a year, looking at what made a particular molecule selective to the target versus another molecule that wasn’t. These insights led them to determine the best compounds to make and test in the lab and, ultimately, they found that RLY-4008 was the most effective.
Promising results so far
Relay began phase 1-2 trials in 2020 and will continue until 2024. Preliminary results showed that, in the 17 patients taking a 70 mg dose of RLY-4008, the drug worked to shrink tumors in 88 percent of patients. This was a significant increase compared to other FGFR inhibitors. For instance, Futibatinib, which recently got FDA approval, had a response rate of only 42 percent.
Across all dose levels, RLY-4008 shrank tumors by 63 percent in 38 patients. In more good news, the drug didn’t elevate their phosphate levels, which suggests that it could be taken without increasing patients’ risk for kidney disease.
“Objectively, this is pretty remarkable,” says Elemento. “In a small patient study, you have a molecule that is able to shrink tumors in such a high fraction of patients. It is unusual to see such good results in a phase 1-2 trial.”
A simulated future
The research team is continuing to use molecular dynamic simulations to develop other new drug, such as one that is being studied in patients with solid tumors and breast cancer.
As for their bile duct cancer drug, RLY-4008, Relay plans by 2024 to have tested it in around 440 patients. “The mature results of the phase 1-2 trial are highly anticipated,” says Goyal, the principal investigator of the trial.
Sameek Roychowdhury, an oncologist and associate professor of internal medicine at Ohio State University, highlights the need for caution. “This has early signs of benefit, but we will look forward to seeing longer term results for benefit and side effect profiles. We need to think a few more steps ahead - these treatments are like the ’Whack-a-Mole game’ where cancer finds a way to become resistant to each subsequent drug.”
“I think the issue is going to be how durable are the responses to the drug and what are the mechanisms of resistance,” says Raymond Wadlow, an oncologist at the Inova Medical Group who specializes in gastrointestinal and haematological cancer. “But the results look promising. It is a much more selective inhibitor of the FGFR protein and less toxic. It’s been an exciting development.”