Deepfakes: Understanding the Technology and Its Implications

Janice_Lin

Everyone knows artificial intelligence is the up and coming, hottest topic of 2023. AI is a decades-old idea, but OpenAI jumpstarted the conversations in late 2022 and caused a global frenzy. It’s important to know that while artificial intelligence is breakthrough technology and revolutionary, there is also a darker side of technology that isn’t included in everyday media. A deepfake is something formed by artificial intelligence that could create numerous consequences that could cause harm, complications, and even law-related problems.

What is a deepfake?  

A deepfake is a form of synthetic media that uses artificial intelligence to replace an existing image or video of a person’s likeness with another person. The technology isn’t new; modern cinema uses CGI to create effects that aren't seen in real life. Deepfakes, however, use extensive artificial intelligence and machine learning to create and manipulate visual and audio content that can easily deceive. To put it plainly, deepfakes can create realistic-looking fake videos at the expense of someone else.

How are deepfakes made? Deepfake technology stems from artificial intelligence. However, there are several ways to create a deepfake video or image. One of the most common ways for deepfake examples to be made is to rely on the use of deep neural networks involving autoencoders, which is a technology that creates face swaps. A main video is first selected, and then a collection of videos or images of the person will be used to insert into the main video. The autoencoder is something that finds the similarities between two faces and decompresses them to become one image or video.

Another type of machine-learning technology to create deepfakes is GANs, or Generative Adversarial Networks. The technology detects and improves the flaws within the deepfake video, making it harder for decoders to spot them. For GANs, the two deep learning algorithms generate several images, and by running loops, a realistic image of completely unreal people will eventually appear.

What is deepfake technology and what are they used for?  

Deepfake technology and deepfake AI are made using autoencoders or generative adversarial networks. Most deepfake videos are made on high-end desktop computers with powerful graphic tools and applications. The expertise to touch upon glitches in videos or images also takes time and strong undertanding of computing tools.

There’s definitely dangerous and malicious power that comes with the technology to create realistic images and videos of other people. Most deepfake examples come from adult deepfakes. Pornography made up 96% of deepfake videos found online in 2019, according to a report by Deeptrace. One deepfake video could be a major threat to actresses, celebrities, and even common folk, damaging their reputation and causing harm. Aside from synthetic pornography, there are also several other applications of deepfake videos that have been applied in real life right under our eyes.

• Political implications. Deepfake videos have been infamous in politics. These are videos specifically created to mislead and manipulate viewers and the public. You might’ve seen a video of Donald Trump calling out Belgium to withdraw from the Paris climate agreement, or Barack Obama mocking a politician of the opposing political party. The political implications surrounding deepfake videos could result in international tension and unnecessary issues.
• Social implications. As mentioned above, adult deepfakes take up the majority of real-life applications. When applied to non-celebrities, there could be severe mental, emotional, and physical harm caused to people who are a victim of deepfake technology. They could lead to exploitation, intimidation, and personal sabotage of individuals. It is a severe unethical imbalance in society and has opened the conversation of legal and regulatory responses to counter deepfake technology.
• Economic implications. Anything’s possible in deepfake technology. For example, if a viral video of Elon Musk doing drugs were going around, stock prices for Tesla and his other companies would most likely drop. Deepfake videos have the power to control consumerism and capitalism. A fabricated video could potentially shift stock prices, cause international tension, and sway voters.

Are they only used in videos and pictures? 

Deepfake technology is not only used in videos and images. There are fast-growing lists of applications in which the technology can be implemented. Digital audio, for example, can now be created in just a few hours to imitate a person talking. This can be used to commit fraud or scams.

How can we spot a deepfake? 

We’ve gone through how to make a deepfake video. But how are they spotted? In the past, it was easy to spot a deepfake video from abnormal blinking and patchy voices or video quality. Sometimes, the angles of the face could appear unnatural and the edges soft, since there wasn't enough video footage to produce high-quality replicas of the person.

There are also other signs to detect a deepfake video. Lighting might be unnatural and inconsistent, and details of finer objects and angles, like jewelry, might be unrealistic. These are some of the major indicators that give away deepfake videos.

What is the future of deepfakes? 

Deepfakes will become more realistic as time goes on and technology improves. A number of companies and even countries have started taking up legal and regulatory measures against deepfake technology. Facebook even hosted a seminar to encourage new startups to detect deepfake and manipulated digital content. There are several initiatives to combat malicious content and fake news, and developing platforms to detect deepfakes are emerging. Artificial intelligence may be the reason why deepfakes became possible, but it is also the solution to combating deepfake content. Blockchain ledger systems could also hold tamper-proof videos and images, and trace them back to their origins so their record can be fact-checked.

While the majority of deepfake technology has insidious intent, not all are. Some deepfake technology developed the power to recreate videos, images, and audio of the deceased, allowing their loved ones to recreate memories with them. Other forms of entertainment use deepfake AI to replicate famous artists in social campaigns or to improve dubbing on foreign films. The point is, deepfake technology and artificial intelligence can be put to good use as well; the most important thing is to properly educate and inform ourselves.