Artificial intelligence-generated videos, commonly known as "deepfakes," are proliferating across digital platforms at an unprecedented pace. These sophisticated manipulations employ cutting-edge computer graphics and audio processing to convincingly replicate human speech and behavior, creating realistic yet entirely fabricated content that threatens to distort our perception of reality and propagate dangerous misinformation. Researchers worldwide have expressed grave concerns about deepfakes' potential to influence electoral processes, notably highlighting their capacity to manipulate American voters during the 2020 elections.
As technology corporations scramble to develop detection mechanisms and policymakers explore regulatory frameworks, an innovative team of artists and computer scientists at MIT's Center for Advanced Virtuality has conceived an educational art installation designed to empower citizens with the critical thinking skills necessary to distinguish authentic content from AI-generated fabrications.
"Computer-based misinformation represents a formidable global challenge," explains Fox Harrell, MIT professor of digital media and artificial intelligence and director of the Center for Advanced Virtuality. "Our mission focuses on substantially enhancing public media literacy. We're committed to leveraging artificial intelligence not for deception, but for promoting truth and transparency. We're thrilled to welcome talented professionals like our new XR Creative Director Francesca Panetta to advance this crucial mission."
Panetta, co-directing "In Event of Moon Disaster" with Halsey Burgund from MIT's Open Documentary Lab, shares: "We aspire to cultivate critical awareness among audiences. We want people to recognize the capabilities of contemporary technology, examine their own vulnerability to manipulation, and develop healthy skepticism toward media content as we navigate an increasingly complex information landscape where truth itself becomes contested."
The "In Event of Moon Disaster" installation, premiering at the International Documentary Festival Amsterdam, reimagines the historic moon landing narrative. Visitors step into a meticulously reconstructed 1960s living room, surrounded by vintage furniture and three screens featuring period-appropriate televisions. These displays present carefully curated NASA footage, taking viewers on a journey from launch to lunar landing. The centerpiece features a deepfake of President Richard Nixon delivering a contingency speech—authored by speechwriter Bill Safire—that would have been broadcast had the Apollo 11 astronauts been unable to return to Earth.
To create this poignant alternative history, the team employed advanced deep learning techniques alongside a professional voice actor to synthesize Richard Nixon's distinctive speaking patterns, collaborating with Ukrainian-based company Respeecher. They partnered with Israeli firm Canny AI to implement video dialogue replacement technology, meticulously studying and replicating Nixon's facial movements and lip synchronization. The resulting video demonstrates the disturbingly convincing capabilities of contemporary deepfake technology.
The researchers strategically selected this historical moment for several compelling reasons: space exploration enjoys universal appeal, ensuring broad engagement; the subject remains apolitical, avoiding unnecessary polarization; and since the 1969 moon landing represents a widely accepted historical fact, the fabricated elements become immediately apparent to viewers.
Enhancing its educational value, "In Event of Moon Disaster" transparently showcases the capabilities and limitations of current AI technology while explicitly aiming to increase public awareness and detection abilities regarding deepfake content. The exhibition features specially created newspapers detailing the installation's creation process, practical guidance for identifying synthetic media, and updates on the latest algorithmic detection research. Visitors are encouraged to take these educational materials home.
"Our objective was to employ the most sophisticated artificial intelligence techniques available to create the most convincing deepfake possible—and then explicitly demonstrate its artificial nature, explaining exactly how and why we created it," explains Burgund.
While the physical installation debuted in Amsterdam during November 2019, the team is developing an accessible web-based version scheduled for launch in spring 2020, extending this crucial educational initiative to a global audience.
