About three years ago one “Synthesizing Obama” program took the internet by storm as the former president was seen mouthing words that clearly contrasted his political stand and years of national image. On a closer view, it was established that the audio-video had been doctored to sync, a genius of the deepfake video technology.
The incident may have generated quite an impact because netizens haven’t stopped trying hands at the expensive technology ever since. Recently, Forbes reported that the deepfake videos election is one of the largest threats this year, as the proliferation of such creators has witnessed a 20 times growth within a year between 2019 – 2020.
That such videos can make or break a political leader’s campaign with doctored deepfake propaganda, is known and feared by most, especially media outlets who would most likely be at the forefront to take the fall of spreading misinformation and controversy. It is suggested that the elections have a lot to do with researchers catalyzing the process of developing extremely accurate deepfake detectors as a preparation for the upcoming expected wave of such propaganda.
In light of these events, what really lies behind the technology, how and why is it thriving despite being a threat to transparency and privacy, and what measures are being taken to counter its negative potential.
Deepfake Technology: What Lies Behind The Fake Face?
Deepfake videos are forms of synthetic media as they call it, which is different from the organic shots or original subjects that are legitimately produced.
The name practically gives away the technology behind — deep learning — a branch of artificial intelligence. Behind the fake faces that we see, lies an intricate neural network working on an encoder and a decoder that is capable to synergize and replicate facial movements to match the desired output.
It has been used as a tool to develop ways to enhance videos for at least a couple of decades now, however, recent controversies and subsequent commercialization of the technology in the form of FakeApp has sparked the trend of amateur production and miscellaneous deepfake video creation.
Contemporary Uses and Misuses Of Deepfake Videos
→ Replacing actors with stuntmen in shots involving dangerous stunts.
→ Minimizing differences between real actors and body doubles.
→ Spoof videos using face swaps.
→ Social media filters to replace or modify certain facial features.
→ Placing influential people in controversial scenarios and presenting it as video evidence.
→ Spreading misinformation in the face of credible experts.
→ Misleading and displacing masses to believe a viral lie.
Deepfake Videos And Reality: The Diminishing Gap
The different ways in which deepfake videos can be misused is terrifying, to say the least. The repercussions of such distorted sources of information and presentation can be adverse and have the potential to make or break a person’s reputation.
If that is not bad enough, the gap between the products of the deepfake video technology and that of authentic videos is diminishing, and there are three pivotal factors that are making the scenario even more problematic.
Deepfake Videos Are Legal
That’s right. Even though the technology can be used to make something outright offensive and traumatizing like face-swapped pornography, the perpetrators can only be called so if a copyright or defamation case is filed by the victim. And that only holds ground in cases of doctored pornography.
For laymen and celebrities that have been wronged in ways that can be categorized as spoof, memes, or graphic dark humor, there is no way to get the synthetic video creators penalized. Considering that no legislature has pointed fingers towards these media creators for years, the technology has been gaining widespread popularity and these creators getting uninhibited encouragement.
Deepfake Videos Are Getting Closer To Reality
There existed a time when veterans from the media industry relied on their experience to differentiate what was real and what was fake, simply because the differences were recognizable and easily visible. These differences involved basic eye blinking, movements, facial gestures, or even body language.
With years, however, the technology and creativity of such content creators have exceeded the capacity of the expert’s eye. The deepfake videos of today look like a stark replica of the original ones, and the similarities are terrifying.
Lack Of Accessible Tools To Detect Deepfake Videos
Perhaps the most grave of the problems is that public and corporate systems still lack efficient tools to detect and distinguish deepfake videos from real ones. The few tools that we have managed to develop over the years to verify video evidence and sources of information are no longer as reliable and can pass an advanced doctored deepfake video as a real one. Needless to say, this can overturn the public, if not national sentiments about influential people for the worse.
Deepfake Videos Detection and Verification: How Far Have We Come?
The above-discussed problems have led to the global realization that we need an even more efficient technology to cut through the emerging menaces that the existing deepfake video technology is making. Not only that, the resulting technology needs to be accessible and affordable enough to media outlets at the least because they hold the power to influence and mobilize masses across the world.
Progress in research and findings along the lines of these conditions have recently been making headlines. These indicate that there certainly is a way out to minutely detect the anomalies of deep fake videos to distinguish them from real ones,
AI vs AI: Researchers Prepare For A Show-Down
Quite recently, Microsoft unveiled its authentication tool that detects deepfakes and face swaps claiming that the accuracy is unmatched and the software can help combat misinformation ahead of the elections in November. Although the algorithm was only found to be highly effective with images and not to the par for footage, this attempt created an impressive stir, leading to several researchers coming forward with their findings in this area. The general agreement and consensus are that the solution lies in the problem, that is, an AI-based deepfake technology can only be countered and verified by AI. The race to develop AI algorithms that can detect the swaps done by AI has escalated and the future looks good, even though uncertain.
Coming Soon, But Not Soon Enough
It is common knowledge that the time around the American elections is watch-dogged by public and corporate systems not just in the country but worldwide. The reason might be the US elections’ immediate impact on global stock markets, and the succeeding turn of events that will determine the diplomatic and trading relations between the superpower and other nations. However, with the elections already staring right into our eyes, there’s little hope to have the technology armor fake videos this time around.
However, researchers are hopeful for the future, and perhaps the interest that social and technology giants Facebook and Microsoft have been taking in developing the technology, a cutting-edge and cutting-deepfake solution may be closer than we expect it to be.
please contact me for more info
