The growing use of deep fake videos in recent years has made it possible to create fake videos and weaponize them to greater effect in ...
The growing use of deep fake videos in recent years has made it possible to create fake videos and weaponize them to greater effect in order to misinform, destabilize, slander, defame, and conduct all sorts of malicious activities that cause a lot of harm and social division. With the exponential improvement of artificial intelligence and machine learning, deep fakes will soon take a mind of their own by distorting facts and rewriting history.
By definition, the term is a portmanteau of "deep learning" and "fake" as this type of synthetic media is basically a digital mask of someone else's likeness with the aim to deceive and misinform the general public. The main machine learning methods used to create deep fakes are based on deep learning and involve training generative neural network architectures, such as autoencoders or generative adversarial networks (GANs).
Perhaps, the earliest form of deepfake is the practice of so-called "revenge porn." It is a type of manipulated porn video generated by using deep fake technology to mimic people performing sex acts they never actually did. Instead, they use existing porn videos and replace the faces of other people, often involving female celebrities (most notably Daisy Ridley and Scarlett Johansson), without their consent. Such videos have increasingly become more sophisticated so that they may look as if it's real. Soon, such deepfake videos have moved away from porn and become more political with videos showing public figures saying something controversial and inflammatory that they never actually said.
Memes have sprouted like mushrooms when social media and Internet trolling have become mainstream. Fan-made videos shared on Youtube and Facebook eventually featured Holywood actors like Nicholas Cage and Tom Cruise hilariously appearing in countless viral videos appearing in films they never starred in.
With the advent of sophisticated artificial intelligence coupled with 3D modelling, voice recognition & generation, and machine learning systems, it is now possible to create videos of ex-President Barack Obama saying something he never said without the use of Holywood special effects. In fact, such technology has evolved that it is no longer just a digital mask super-imposed in an existing video or digital puppetry controlling a target person's digital avatar. AI can now create new videos from scratch and even generate their images of a target personality through machine learning.
With the rise of totalitarian states in the Interwar Years, Nazi Germany, Fascist Italy, and the Soviet Union have produced their own brand of propaganda films. The United States even joined the bandwagon with Holywood producing their parody of Adolf Hitler in Charlie Chaplin's "The Dictator." (1940) Although not considered deep fake by today's standards, Hitler eventually became the most portrayed public figure in the history of propaganda films.
After the war and the whole stretch of the Cold War, the manipulation of films with certain political objectives has intensified through the careful incorporation of subliminal messages
In perhaps the earliest proper description of what we know as deep fake, the 1986 December issue of Analog magazine under Jack Wodhams' novelette "Picaper" mentioned digitally-enhanced or generated videos produced by skilled hackers serving unscrupulous lawyers and political figures. He called such fabricated videos as picaper as the title suggests. It is also known as mimepic, which means image animation based on "the information from the presented image, and copied through choices from an infinite number of variables that a program might supply." Wodhams closed it off with a sobering conclusion that "the old idea that pictures do not lie is going to have to undergo drastic revision."
It is ironic that the academic world laid the groundwork for developing the building blocks of deepfake technology. In 1997, the Video Rewrite program has allowed users to full automate facial reanimation by leveraging machine learning to make the connections between the sounds produced by a video's subject and the shape of the subject's face.
It was in December 2017 when deep fakes started to gain attention after an article was published on AI-manipulated porn that featured actress Gal Gadot. It was posted by a prolific Reddit user named "deep fakes," who later published a series of fake celebrity porn videos. From then on, it took a life of its own as these were made using open-source machine learning tools like TensorFlow, which Google makes freely available to anyone interested in machine learning.
By 2018, the development of software and mobile apps like FakeApp and Momo has made it possible for ordinary people to create their own deep fake.
By definition, the term is a portmanteau of "deep learning" and "fake" as this type of synthetic media is basically a digital mask of someone else's likeness with the aim to deceive and misinform the general public. The main machine learning methods used to create deep fakes are based on deep learning and involve training generative neural network architectures, such as autoencoders or generative adversarial networks (GANs).
Emergence of Deepfake
Deepfake may sound like a modern phenomenon in this social media age but disinformation has always been practiced since time immemorial. Depending on which perspective you believe, governments and regimes have unleashed propaganda films and movies by distorting facts, twisting the narrative, and vilifying a common target to justify their grip on power. With the democratization of information with the growth of the Internet and social media, disinformation and propaganda have been rebranded as fake news and deep fakes.Perhaps, the earliest form of deepfake is the practice of so-called "revenge porn." It is a type of manipulated porn video generated by using deep fake technology to mimic people performing sex acts they never actually did. Instead, they use existing porn videos and replace the faces of other people, often involving female celebrities (most notably Daisy Ridley and Scarlett Johansson), without their consent. Such videos have increasingly become more sophisticated so that they may look as if it's real. Soon, such deepfake videos have moved away from porn and become more political with videos showing public figures saying something controversial and inflammatory that they never actually said.
Memes have sprouted like mushrooms when social media and Internet trolling have become mainstream. Fan-made videos shared on Youtube and Facebook eventually featured Holywood actors like Nicholas Cage and Tom Cruise hilariously appearing in countless viral videos appearing in films they never starred in.
With the advent of sophisticated artificial intelligence coupled with 3D modelling, voice recognition & generation, and machine learning systems, it is now possible to create videos of ex-President Barack Obama saying something he never said without the use of Holywood special effects. In fact, such technology has evolved that it is no longer just a digital mask super-imposed in an existing video or digital puppetry controlling a target person's digital avatar. AI can now create new videos from scratch and even generate their images of a target personality through machine learning.
Historical Precedents
Although deep fakes are considered to have gained prominence in 2017, their roots began way back during the early years of film and moviemaking. By the turn of the century, propaganda films were made to "re-enact" certain historical events and portray prominent figures in a good light. One of the earliest propaganda films that portrayed a current leader in a dramatic way was "October" (1927) by Sergei Eisenstein. Lenin was portrayed by an unknown actor when he declared that the government was overthrown as the revolutionaries stormed the Winter Palace. There was no documentary evidence to prove that the event even happened. |
| October was perhaps the most prominent film to have portrayed a real-world leader at that time |
After the war and the whole stretch of the Cold War, the manipulation of films with certain political objectives has intensified through the careful incorporation of subliminal messages
In perhaps the earliest proper description of what we know as deep fake, the 1986 December issue of Analog magazine under Jack Wodhams' novelette "Picaper" mentioned digitally-enhanced or generated videos produced by skilled hackers serving unscrupulous lawyers and political figures. He called such fabricated videos as picaper as the title suggests. It is also known as mimepic, which means image animation based on "the information from the presented image, and copied through choices from an infinite number of variables that a program might supply." Wodhams closed it off with a sobering conclusion that "the old idea that pictures do not lie is going to have to undergo drastic revision."
It is ironic that the academic world laid the groundwork for developing the building blocks of deepfake technology. In 1997, the Video Rewrite program has allowed users to full automate facial reanimation by leveraging machine learning to make the connections between the sounds produced by a video's subject and the shape of the subject's face.
It was in December 2017 when deep fakes started to gain attention after an article was published on AI-manipulated porn that featured actress Gal Gadot. It was posted by a prolific Reddit user named "deep fakes," who later published a series of fake celebrity porn videos. From then on, it took a life of its own as these were made using open-source machine learning tools like TensorFlow, which Google makes freely available to anyone interested in machine learning.
By 2018, the development of software and mobile apps like FakeApp and Momo has made it possible for ordinary people to create their own deep fake.
No comments