The Future of Leadership/Understanding the Nature of Deepfakes

What is a Deepfake and How is it Created?

A Deepfake is an AI generated overlay used to swap faces or other items in media. The term Deepfake is a portmanteau of the “Deep Learning” and “Fake” referring to the usual nature of the altered media being used to “fake” a video or audio media of an individual performing actions. One of the more common methods of creating a deepfake is to use a deep learning neural network. To create the deepfake two sets of source data are needed. The first source, which will be referred to hereafter as the “Source Individual’ is photographic, audio, and video media of the individual to be faked. The second source which will be referred to as the “Source Media” is the collected data of the actions that the source individual will be faked to perform. Both sets of data are fed into a neural network encoder that breaks down the movements, audio, and facial data into compressed information. Then, the source individual data is decoded by the source media decoder, which overlays the information from the source individual over that of the source media. This creates the faked media in which the source individual appears to perform actions.

The Ethics of Deepfakes

Are Deepfakes moral? Many would argue they are not moral. The moral grounds for deepfakes can largely be boiled down to the question of to which degree an individual may be considered to own their own appearance, and furthermore their identity. Movie studios have begun to make use of deepfakes, with a high-profile example being that of Star Wars, which used deepfake technology to digitally recreate Cary Fisher after her passing, and to portray a younger version of Luke Skywalker in media despite the fact that Mark Hamil has aged far beyond the age portrayed in the media. In the case of Mark Hamil, his participation in the creation of, and authorization of the use of the deepfake can be viewed to allow the ethical use of his visual identity. In the case of Cary Fisher, who had passed away prior to the creation of the deepfakes, the issue is murky. At the core, the issue of identity comes into question through the use of deepfakes. Were identity to be fully self-contained, that is to say, completely under the control of the self, then deepfakes would present little moral issue. However, a part of identity can be viewed as interconnected with the views and opinions of others towards us. One may consider oneself a moral person, however if said individual has not behaved morally and is viewed as immoral by their community, then the identity of a moral individual cannot truly be said to belong to them. This is an ethical problem with deepfakes. They represent a means by which the identity of an individual can be tarnished or damaged through no action of that individual.

Ethical Opportunities of Deepfakes

As presented in brief in the paragraph above, the ethical use of deepfakes in media represents a massive opportunity for media creation companies. A beloved actor may license out features, visual or audio, to a movie studio, allowing them to remain a part of the body of work long after age and potentially even death renders them incapable of active participation. A further opportunity created by deepfakes is that of historical learning. When science is able to recreate an estimation of the appearance, voice, and other data from a historical figure, then history can be brought to life. The opportunity to engage students by having them learn “directly” from a historical figure represents an opportunity for far greater engagement than many students would receive from reading a textbook or watching a historian pontificate on a topic. A further opportunity found by individuals online is the use of purposely and openly fake “avatars” in order to create a “virtual youtuber” personality known as a V-tuber. These V-tubers have gained popularity and become a form of job for not only those who create them, but those who portray themselves through the medium.

The Threat Posed by Deepfakes.

While deepfakes represent opportunity for some industries, they represent a far greater threat to others. The primary threat comes from misinformation and disinformation campaigns using Deepfakes to portray false messages as true. With the use of deepfakes, anyone can appear to say anything. From former President Barack Obama engaging in a profanity laden rant against Former President Donald Trump, to a video message appearing to be from Ukrainian President Volodomyr Zelinski instructing Ukrainian citizens to surrender and embrace the rulership of Russia, several issues immediately become apparent. For any career which requires a view of integrity, deepfakes represent a method by which bad actors can immediately destroy credibility. A recent scandal involving deepfaked pornography which allowed individuals to order deepfakes of any individual performing pornographic actions raised questions to the legality of created materials. In some countries, such creations are already illegal under “Revenge Porn” laws. However, in many countries there is no such legal recourse. While it may be amusing to see Joseph Stalin and Adolph Hitler singing pop music from the 90’s across from each other, it is far less amusing to the employees of a German power company who recently transferred a large sum of money on the orders of the company’s CEO, only to find that the video call had been a deepfake.

Potential future effects of Deepfake technology.

As deepfake technology improves it becomes harder and harder to determine what is real and what is fiction. Only a few months ago the adage was “Look at the eyes'' because the generation of neural networks at the time struggled to handle involuntary reflexes like blinking or sneezing, the rapidity and relative complexity of the actions causing the deepfake technology to struggle to mimic them successfully. No sooner had the limitation been pointed out however, than the neural networks were retrained and adapted, and further adjustments made in turn to combat each weakness discovered. Skin tone, blurring around where the faked media is joined with the source media, and hair have all been improved upon to blur the line between fact and fiction.

Assessing potential methods of combating Deepfakes.

A broad reaching alliance of myriad companies have all joined forces to tackle the issue of deepfakes. Experts from companies like Microsoft, Adobe, Apple, news agencies, governments, and more have all stepped into the ring in order to combat the emerging dangers represented by deepfake technology. Their current “best idea” is to tag every image on the internet with a stamp that allows viewers to trace the “genealogy” of that piece of media through each rendition or iteration, and to see what alterations have been done to that piece of media. This unfortunately is not likely to work, as any actor sophisticated enough to produce a convincing and high quality deepfake is likely to be more than capable of producing an altered or falsified “genealogy” purporting the media to be real. Further steps in the hope of combating deepfakes comes from the idea of using a randomly generated alphanumeric sequence that will be required to be read, in the hopes that the deepfake cannot adapt fast enough.