The deepfakes of the 12 months went mainstream
In 2018, Sam Cole, a reporter at Motherboard, discovered a new and unsettling corner of the internet. A Reddit user called "Deepfakes" posted non-consensual fake porn videos using an AI algorithm to turn celebrities' faces into real porn. Cole alerted the phenomenon when the technology was about to explode. A year later, deepfake porn had spread well beyond Reddit, with easily accessible apps that any woman photographed could use to "undress".
Since then, deepfakes have had a bad rap, and rightly so. The vast majority of them are still used for fake pornography. An investigative journalist was severely harassed and temporarily silenced by such activities, and more recently a poet and writer was scared and ashamed. There is also a risk that political deepfakes will generate compelling bogus news that can wreak havoc in unstable political environments.
However, as the algorithms for manipulating and synthesizing media have become more powerful, they have also spawned positive uses – as well as some that are humorous or mundane. Here is a roundup of some of our favorites in a rough chronological order and why we believe they are a sign of the future.
In June, Welcome to Chechyna, an investigative film into the persecution of LGBTQ people in the Russian Republic, was the first documentary to use deepfakes to protect the identity of its subjects. The activists who fought the persecution and served as the main characters in the story lived in hiding so as not to be tortured or killed. After director David France explored many methods to hide their identity, he decided to give them in-depth "covers". He asked other LGBTQ activists from around the world to give their faces, which were then grafted onto people's faces in his film. Technology enabled France to preserve the integrity of the facial expressions of its subjects, and therefore their pain, fear and humanity. In total, the film shielded 23 people and did pioneering work for a new form of whistleblower protection.
PANETTA AND BURGUNDY
In July, two MIT researchers, Francesca Panetta and Halsey Burgundy, published a project to create an alternate story of the 1969 Apollo moon landing. In the event of a lunar disaster, the speech used will be that President Richard Nixon would have made if the momentous opportunity had not gone according to plan. The researchers worked with two different deep fake audio and video companies and hired an actor to provide the "baseline" service. Then they ran his voice and face through the two types of software and sewed them together into a final deepfake Nixon.
While this project shows how deepfakes can create powerful alternative stories, another suggests how deepfakes can bring real history to life. In February, Time Magazine recreated the Martin Luther King Jr. March in Washington for virtual reality to immerse viewers in the scene. The project didn't use deepfake technology, but Chinese tech giant Tencent later quoted it in a white paper about its plans for AI, saying deepfakes could be used for similar purposes in the future.
MS TECH | NEURIPS (TRAINING KIT); HAO (FRIENDLY PERMISSION)
In late summer, the Memersphere got its hands on easy-to-make deepfakes and released the results into the digital universe. In particular, a viral meme called "Baka Mitai" (pictured above) rose rapidly as people learned to use the technology to make their own versions. The specific algorithm that drives the madness comes from a 2019 research report that allowed a user to animate a photo of one person's face with a video of another person's face. The effect isn't high quality in any way, but it definitely creates quality fun. The phenomenon is not entirely surprising; Play and parody have been a driving force in popularizing deepfakes and other media manipulation tools. For this reason, some experts stress the need for guard rails to prevent satire from becoming abuse.