They say that seeing is believing. But what happens when you cannot believe what you see?
“Reality is one of the few words which means nothing without quotes.”
Vladimir Nakobov
In 2018, sp.a, a Belgian political party posted a one-minute video of the former US president, Donald Trump. The video was a classic case of good intentions gone wrong: it nearly triggered a diplomatic crisis. The catch? The video wasn’t real.
The video in question featured Mr. Trump making statements such as, “Belgium, don’t be a hypocrite, withdraw from the Paris Climate Agreement”. The video ends with him declaring, “We all know climate change is fake, just like this video.” However, the volume of the clip drops drastically at this point, with these lines not being added in the Dutch subtitles.
The aim of the video? To create increased awareness on climate change, and draw the attention of the Belgian government towards it. However, it led to outrage on social media, with many Belgians criticizing the gall of a foreign president to instruct the government of another country. And that’s just the tip of the iceberg for the nefarious potential of deepfakes.
A deep dive into deepfakes
In simple terms, deepfakes are manipulated videos that replace the original person in the video with the likeness of another. In other words, they are the video equivalent of a photoshopped image, created by feeding multiple images of a person to a computer algorithm.
The ethical considerations of deepfakes have also been debated, with some arguing that deepfakes are a net positive to society. However, the uses of deepfakes are multifaceted - they can range from innocent, to disturbing. For instance, The Dalí Museum in St. Petersburg has created a deepfake of Salvador Dalíthat interacts with visitors and even takes selfies with them, forever imprinting a memory of the man himself in their minds.
But there’s a flip side.
Somewhere in the dark, grisly corners of the internet lie thousands of deepfaked pornographic videos of celebrities, journalists, and even normal people. These videos, which are almost always non-consensual, constitute 96% percent of all deepfaked videos, according to DeepTrace Labs.
What’s more concerning, however, is the lack of expertise needed to create one. Any person with access to a computer and an internet connection can create them, and their victims are likely to be those with a large public presence. With an increase in the number of deepfake apps, which range from DeepFace Lab, Zao, FaceApp (yes, the FaceApp), and FaceSwap, it’s now much easier to create them, and much harder to classify them as fake. AI generated photos and videos can be extremely convincing if used correctly (see below), and their past uses have been far from good.
Figure: An example of an AI generated face, from the site https://thispersondoesnotexist.com/
Deepfakes and the distortion of reality
A technology that could show people saying and doing things they’ve never done could wreak havoc on society. Imagine the impact of the videos of a politician making career-ruining statements, a CEO stating that their company’s main objective is deceiving their users, a journalist being non-consensually cast in pornographic videos, or a beauty mogul proudly letting us know that her consumers’ data has made her rich. While these scenarios may seem far-fetched, they’re not far from reality, with the people in question being Barack Obama, Mark Zuckerberg, Rana Ayyub, and Kim Kardashian respectively.
Though the people mentioned above are celebrities, the dangers of deepfakes are slowly seeping into the realm of common people as well in the form of revenge porn, defamatory videos, and blackmail. In one peculiar instance, a woman from Philadelphia used deepfakes to harass a rival cheerleading squad, by creating videos where the girls were shown nude, drinking, and vaping.
Deepfakes can also be used for fraud and extortion. For example, in March of 2019, the CEO of a UK-based company was scammed into transferring £243,000, by whom he thought was a superior at his parent company. Lastly, and most importantly, deepfakes could be used to manipulate elections, resulting in the dismantling of democracies. For instance, a strategically timed deepfake of a political rival before an election could result in a drastic change in public sentiment.
In an age where we consume exorbitant quantities of information on a daily basis, it becomes hard to filter out what’s real and unbiased, and what’s fake and manipulated. It becomes even harder to identify deepfakes, especially as the technology keeps evolving and improving. For instance, the first few deepfakes that came out were extremely grainy and blurry, but more recent ones have become extremely detailed.
Figure: A still from a deepfake of the American actor, Tom Cruise, posted on the TikTok account @deeptomcruise
The legal consequences of deepfakes are blurry. However, many countries are catching up, with the USA passing the Deepfakes Accountability Act in 2019. Though India does not have any laws which explicitly ban deepfakes, however, a person can seek relief under other laws already in place, such as the Information Technology Act 2000, The Copyright Act 1957, and Sec. 499 of the IPC.
Solutions and Strategies:
The task of identifying deepfakes is extremely difficult, due to the technology quickly improving with every passing day. As of 2021, deepfakes are a cybersecurity threat, with many businesses, financial institutions and even governments scrambling to counter them. Certain methods that have been devised to counter deepfakes are as follows:
Deepfakes could perhaps be countered by its very creator, artificial intelligence. Many startups have tried to develop technologies centered around deepfake detection, which include companies like Sensity and Minerva. In 2019, FaceBook announced the Deepfake Detection Challenge, which, as the name suggests, encouraged participants to develop tools to identify deepfakes.
Ensuring stricter regulation against deepfakes on social media platforms, which could help stop their spread entirely if done right.
Legal measures to counter deepfakes could also be imposed. Many countries do not have laws prohibiting the creation and the spread of deepfakes for unethical reasons.
Lastly, the importance of media literacy can’t be overemphasized, especially when we live in a world where we consume large amounts of data every single day. It is imperative for us to educate all generations, both old and young, about the realities of the online world.
Regardless of the solutions we implement, deepfakes are here to stay. It's our responsibility to make sure that’s a good thing.
Author’s Note: In order to get a glimpse of how convincing AI can be, you can check https://thispersondoesnotexist.com/ for hyper realistic AI generated images of people. You can also test your deepfake detection skills at https://www.whichfaceisreal.com/index.php.
References:
Commentaires