Washington – On August 26, during a workshop at the Air Force Military University in Montgomery, Alabama, students switched effortlessly between five languages, including Mandarin and Russian, as a video of President Joe Biden addressing the United Nations was shown.
While Thomas Jefferson and John Quincy Adams were fluent in several languages, Biden, like most US presidents, is known to speak only English.
The video was a piece of synthetic media commonly known as a “deep fake.” Deepfakes, created using machine learning and artificial intelligence, are highly realistic videos that replace a person’s likeness with another person or show them doing something they didn’t do.
And as the technology improves, they become harder to detect.
The workshop, which involved artificial media startup DeepMedia, focused on airing deep simulations and their connection to military applications, as well as introducing the company’s AI technologies to counter potential threats.
Earlier this year, Ukrainian President Volodymyr Zelensky released a deeply fake video of the country calling on its troops to lay down their weapons and surrender in the country’s conflict with Russia. Although the video was quickly removed and taken down, the technology is being used for propaganda and disinformation campaigns.
Rijul Gupta, co-founder of Oakland, California-based DeepMedia, calls video technologies the “next frontier,” citing their low production costs and relatively quick turnaround times.
“Everything can be manipulated in real time,” he said at the workshop.
The company created the Biden video to show that not all deep lies are harmful. DeepMedia says the AI-based universal translation tool could be used to facilitate diplomatic communications, for example.
The translator tool takes a video of someone speaking one language and transforms it so that the subject appears and appears to be speaking a different language in real time. The technology can also be used to suggest linguistic parameters that distinguish deep fake from real video.
Hollywood has been using some profound fakery technology for years, to give old actors new “life” to make dead actors look decades younger. Actor Peter Cushing appeared in “Rogue One: A Star Wars Story” in 2016, two decades after his death.
Still, deep-fake stories about celebrities and other politicians have been widely circulated on social media video platforms, raising fears that man-made media could be used to spread misinformation.
In the year According to a 2021 report from the US Department of Homeland Security, the threat of deep fakes lies not in the technology used to create them, but in people’s natural tendency to believe what they see. Thus, deep lies do not need to be particularly sophisticated or believable to be effective in spreading misinformation, the report noted.
DeepMedia specializes in artificial media discovery and production. In April, the Air Force Research Laboratory announced a partnership with the company to detect deep fakes.
One of the main ways to identify if a video is deeply fake is by analyzing the language used, which reveals differences in pronunciation between the subject of the video and the voice presented.
According to company founder Emma Brown, analyzing videos for those differences takes time, money and effort.
In most cases, the cryptologic linguist conducting the analysis must have a deep understanding of the language and vocabulary, as well as the culture and history of the country. It is not enough, she said, to have a cursory knowledge of the language to spot flaws in deep fakes.
As the military grapples with a major recruiting crisis, attracting the talent needed to fill cryptographic linguist positions poses another challenge. To replace human talent, Brown and Gupta explained that DeepMedia has developed AI tools to replace linguists.
“There are not enough people outside of AIA to do this,” Gupta said.
Catherine Buchaniec is a reporter at C4ISRNET, where she covers artificial intelligence, cyber warfare and emerging technologies.
