It wasn’t long ago that a person could cite a video recording as evidence that something occurred. However, that’s not the case anymore due to the rise of deepfake video. Deepfakes are pictures, audio recordings or video footage that appear so real that there’s no sign of inauthenticity.
It’s easy to imagine how a person could use a deepfake to assert that a world leader or celebrity said something they didn’t. Or, a disgruntled spouse could even create a deepfake video that apparently shows their partner in bed with someone else.
How is Deepfake Video Made?
Deepfake videos became possible due to machine learning models called “generative adversarial networks” (GANs). They work when a generator model trains on a data set and creates fake content.
Another so-called “discriminator model” tries to detect what’s not real. The generative model learns from its own past attempts and uses them to create progressively more realistic content. The discriminator aims to identify what’s new and what’s part of the original training data. When it can no longer tell, the generator has successfully fooled the discriminator.
Why Are Deepfakes Dangerous?
Most of the known deepfakes created so far were done for research purposes or to show people what’s possible and make them more aware. Those instances also explain why people should consider a deepfake video or other kinds of media as dangerous, and not something only created for fun.
1. Deepfake Video Could Become Court Evidence
Byron James, a family law lawyer based in Dubai, says that someone on the opposing side of his client used software to fake a recording of the person making a threat over the phone. The lawyer eventually proved the recording as fake and cautioned that one of the methods for revealing deepfakes involves scrutinizing the original recording.
A deepfake video created for this purpose could be even more of a threat, since a judge or other legal professional could assert that they see a person on screen as well as hear them, making the footage all the more authentic. There is already a long history of issues with the legal system that result in innocent people being jailed.
Fake video could be added to that list, especially if people don’t take care to scrutinize what they see and hear. If deepfakes become more common, the safest tactic to practice may even be believing the footage isn’t authentic until people get evidence that proves otherwise.
2. Deepfakes Could Damage a Person’s Reputation
Companies that specialize in making video presentations for their clients are also affected by the rising amount of fakery in the media. There’s an ongoing struggle to show content in ways that are true and believable. For example, although executives are usually credible sources to tell the story of a brand, the most accurate information often comes from the everyday employees who work for an organization.
One of the reasons why fake news has taken such a hold in today’s society is that people often share the content they see or read without verifying its accuracy. People have already successfully shown how it’s possible to create deepfake videos of people in power uttering things they never said in real life.
The people behind a media-based art project made a fake video of Facebook founder Mark Zuckerberg giving a news interview about Facebook’s power. The creators then successfully posted the clip to Instagram, although they made sure to designate it as art, not a real broadcast. Another notable example of a fake video involved writer-director Jordan Peele creating a deepfake video of Barack Obama seemingly delivering a speech.
If a person wanted to act maliciously towards someone in an authoritative position, getting results might be as straightforward as creating a fake video that damages their reputation. Videos can spread across the internet and reach a global audience in a matter of minutes. That level of reach and speed makes it exceptionally difficult for a company to do damage control in such situations.
3. People Find It Hard to Detect Fake Content
Another worrisome aspect of the deepfake trend is that people have difficulty spotting fake media. Stanford University researchers concluded that 52% of high schoolers believed that a poor-quality video that claimed to show ballot stuffing constituted strong evidence of U.S voter fraud during the 2016 primaries. That video was not a deepfake, but it was from Russia and not the United States.
However, the researchers found the results worrisome, especially since many of the high schoolers in the study will be able to vote for the first time in the 2020 election. Then, concerning older generations, many people aren’t accustomed to thinking that technology is so advanced that it can make a believable fake video, so they assume what they see is real.
A positive factor is that teams are making progress with advanced detectors that find instances of fake videos. For example, UC Berkeley researchers trained software to look at the hyper-precise facial movements people make, such as the rotation of the head during a frown.
The Future Demands Stricter Evaluations of Media
The examples here show why people can no longer necessarily believe what they see or hear. Although this coverage focused on fake videos, there are equally believable examples of images. Media can increase a person’s understanding of an issue and clarify points that may otherwise confuse. However, it’s easier for everyday people to access advanced software that lets them doctor media to convey the things they want.
This shift in the media landscape means people must check facts and verify sources rather than automatically believing what’s in front of them is true. That change will take time, but it’s essential for members of society to understand the risk of deepfakes and do their part to stop their spread.
