Ai + Society — Deepfakes Project
by Jessica Li and Sarah Jang
The effect of deep fakes on society and the potential harm it can do to us.
Introduction
Deep fakes are an ever-impending issue in the world today and for the future. With the advent of new technology, especially in Machine Learning and in AI, there will inevitably be new technologies that will take a shape of their own, opposing the intentions of their original purpose. This position could be said about Deepfakes. Starting with academic research on the generative adversarial network (GAN) in computer vision, Deepfakes have now been used to infest pornography sites with AI-generated actors/actresses, cause identity hoaxes, and spread fake news.
Quite interestingly, Deepfakes were first used in the mainstream in 2017 within the Reddit community where a couple of Reddit users used Nicholas Cage’s face in place of popular pornographic actors/actresses to simulate the actor or actress action in their place. From that, the prevalence of Deepfakes has since become widespread, with Obama, Trump, and Pelosi all being Deepfaked into saying obscene or fraudulent things and uploaded onto social media. There have been many hate crimes and pornographic revenge exploits caused by using the software to fake someone’s actions. However, there are some positives that come out of Deepfakes. Deepfakes can help in creating AI for movies and media, where some situations do not require the actor to risk themselves. For example, in an action movie, the fight sequences can be easily arranged by Deepfaking the main actor’s face onto their professional stunt man/woman’s body.
Overall, there are more negatives than outweigh the good with Deepfakes. And as technology may advance to make our lives more efficient, there will always be a price to pay, and with Deepfakes, it’s our privacy and our identity itself.
Medias
Within this project, we have explored a couple of different avenues that Deepfakes have had a sizable impact: pornography, political, arts, and social media industries
Pornography
A study done by AI firm Deeptrace in September 2019 found that 96% of deep fake videos are pornographic. Most of the time the faces of female actresses are mapped on to porn stars, but profile pictures of women on social media have been edited on these videos as well. Noelle Martin is at the forefront of bringing awareness on deep fakes and proposing laws against image and media abuse. At age 17, Martin received a pornographic deep fake of her from an anonymous email. When she searched the photos of her through Google images, thousands of thumbnails and videos of her face were attached to porn stars. That was the start of an exhausting five year battle of convincing porn sites to pull down her fake videos and convincing the Australian government to criminalize such offenses.
https://www.elle.com/uk/life-and-culture/a30748079/deepfake-porn/
Political
Another very famous example of Deepfakes in media was in 2016 when Jordan Peele deep faked Obama and, with Buzzfeed Publication, sent out a fake video of Obama saying very defamatory and explicit things. The video was used as a PSA to the general public to not trust anything that is on the internet. Because the video had gone viral, people who had never heard of deep fakes were finally exposed to how bad deep fakes can be. Before viewing this video, my young people including myself had never even heard of a deep fake. Unfortunately, even though a prominent celebrity figure had brought this issue to light, unsurprisingly, political deep fakes are still being made: Trump, Pelosi, and Elizabeth Warren deep fakes have all spread through the media. Luckily, deep fakes are not currently prevalent, but going forward the implications are totally viable in creating large disruptions in our political ecosystem and the ways that our government is run.
https://www.youtube.com/watch?v=cQ54GDm1eL0
https://www.youtube.com/watch?v=MoHLM4XMJcQ
Acting/Art
With the rise in fantastical and science fiction movies, there has been a rise in CGI (Computer-Generated Imagery). CGI is not the same as deep fakes, but there are some aspects of CGI that take from deep fakes and vice versa. A huge example of this would be in Fast n’ Furious 7, Maleficent, and in The Curious Case of Benjamin Button. All three of these movies used computer graphics to put in the actor/actresses’ faces into the movie where it was not originally, such as Brad Pitt’s older-version face. Just like deep fakes, the computer is making a new persona out of an older existing person. Especially in Fast n’ Furious 7, the team used older clips of the late Paul Walker, to create a version of Walker into the movies even postmortem. With the growth of deep fakes, it is soon that they will be used in Hollywood to decrease the actor, actress’s costs, and create movies where you can never even tell that no one was there to film that scene.
https://www.youtube.com/watch?v=edG0TF7xpiQ
Social Media
Social media is part of most people’s lives in the 21st century. With there being growing numbers of applications to download and find on the web and more global interactivity between each community, there will always be the newest technology that can spread. And even though it is not explicitly deep fakes, there is something very similar that we all hold possession of because we have these social media apps. These types of “deep fakes” pervade through our Instagram, Facebook, and Snapchats, and that is filters. Filters use neural networks and to overlay a mask on top of what is on the screen. A lot of people love the blurring filters to make their skin smoother or maybe adds a type of style to the video. And recently, there has even been a Snapchat filter that switches males to females and females to males convincingly. There are even articles where people have catfished others from these filters. Although not completely generated by AI, this deep fake-like software are similar and acceptable enough that people will become less and less scared of having legitimate deep fakes will be in our pockets.
Old person Snapchat Filter: https://www.youtube.com/watch?v=4nTqlLvCOG8
Male/Female Snapchat Filter: https://www.youtube.com/watch?v=c2j_kOlS_5w
Conclusion
It is exciting that we have come so far with technology that reality itself can seem to be bent. Some of our most ambitious science-fiction dreams can be realized with the innovations we have today. However, with more innovation, there must be a sacrifice and the sacrifice for the AI-generated facial technology of deepfakes is our identity. Deep fakes are extremely dangerous as it takes away our ability to be ourselves. It takes away our voice, our privacy, and ultimately our wills. So in the future, the world needs to be more aware of deep fakes. We must keep ourselves and our loved ones sharp and safe.