The phenomena of deepfakes is nothing new, what is new is the quality and the speed that they are spread into the world with. Thus, also the impact and the size of the snowball effect they can provoke in society.
What was considered as the first deepfake happened about five years ago, when an anonymous internet user published pornographic videos changing the faces of the women that appeared in it with faces of famous actresses. Back in 2019 the Dutch cyber security company Deeptrace detected around 15.000 deepfakes videos on the internet. Imagine what that number looks like today considering the impressive development of technologies that we’ve experienced in the last years and that is probably even going to increase in the following. But before we dive into numbers, threats and the question if there are even any benefits behind deepfakes, let’s take a closer look at what we actually mean when using this term.
What are deepfakes?
The term “deepfake” is composed of two words, on one hand side “deep”, which alludes to deep learning, the automated learning process of any generative artificial intelligence software. It takes place in a so called Generative Adversarial Network – GAN – where two neuronal networks are combined, a generative one that creates the false image and a discriminator that introduces that image into the real sequence. That’s where the second term comes into the game, “fake”, which clarifies that it refers to false content, images that are not real and sequences that have never happened that way. Thus, a deepfake is an image or a video modified by artificial intelligence that seems to be original and real. There are generally three types of deepfakes:
- Deepface: Refers to a visual manipulation, as these usually show the faces of people because they are easier to create. Deepfaces can appear in different formats such as video, image or even live streamings.
- Deepvoice: Refers to a manipulation of the voice, basically, when words are put into someone’s mouth that have never been said. This kind of deepfake typically appears in the form of voice recordings.
- Deeptext: Also in form of text we can find fakes created by artificial intelligence, such as mailings, posts in social media platforms etc.
Whereas a few years ago it may have been possible to detect a deepfake more or less easily, nowadays it’s hardly impossible as technology has improved significantly and keeps doing so constantly at an impressive pace. On top, by now literally anyone can create a deepfake and spread it in no time through social media.
Why would you do so? Well there are a variety of intentions behind the whole story. Some deepfakes are just created to entertain and are quite harmless. Unless the author’s sense of humor is based on making fun of somebody else maybe. Also for marketing purposes they can be used in a way that doesn’t disadvantage anyone, for example to make campaigns more illustrative showing in an obvious way that it is a false image or declaring so upfront, without trying to spread false information. That’s exactly what happens a lot with deepfakes of celebrities or public or political figures. In these cases deepfakes are often used to desinform or manipulate the public opinion.
Recent examples of deepfakes in different contexts
- Entertainment
Here is one example of a deepfake made to entertain, inviting celebrities to talk about a certain topic sitting together at a table, just without having ever met to do so. As the name of the channel itself indicates that the content shown is a deepfake, the intuition is clearly not to mislead anyone. At the same time it’s impressive how good the fake representations of those celebrities are, including their voices!
- Marketing
The first official commercial case of a deepfake was in 2021, when Bruce Willis officially allowed the Russian telecommunications company Megafon to use his image for a commercial campaign. After some controversial discussions he clarified that he didn’t sell the right to use his appearance forever and in whatever context, but just for this commercial. Again, in this case, as the information about the technology used wasn’t kept a secret, there is no harm.
- Public or political figures
As mentioned before deepfakes are being used more and more in politics as well, even consciousness and on purpose by politicians who want to appear speaking languages they actually don’t, to mention a basic case, but also to show themselves in situations that make them appear more glorious during election campaigns for example. Situations that have actually never happened though. And that’s where we cross a fine line of how we perceive deepfakes.
Concerns and risks related to deepfakes
Something probably all the technological advances have in common is that they can be used the good and the bad way. The problem is, that it’s on us to do so, and on this planet we are many human beings, so the probability that not all of us use technology for the good is quite high. And that’s when disadvantages and concerns appear.
- Mistrust
When political or public figures use deepfakes, voluntarily or not, it inevitably increases mistrust. Fake news has always been spread one way or another, but being able to actually make a person appear and say things, even though it never happened, and make it seem real, brings fake news onto a whole other level. So on one hand, if politicians for example actually embrace this option and use it for their benefit, mistrust increases as nobody can tell anymore what has actually happened knowing that there are false as well as true images out there. On top of that there are the images that politicians did not produce intentionally, considered attacks or cases of abuses, as probably a counter person uses them to make the person appear in a non-beneficial situation. With this panorama out there, between false images for the good, false images for the bad, and yes, believe it or not, also some true images, people will lose more and more trust in digital representations.
- Manipulation and Desinformation
Which leads to the next point, manipulation and desinformation. May it be for economic, political or any other reason, it is clear that the ability to create false images, videos and voice recordings gives any person with the wrong intentions an infinity of possibilities to manipulate information and spread wrong facts around the world, literally. This can influence the course of elections, the finance market, harm international relations and have impacts in literally any other scenario you can imagine.
- Descreditation
This goes in line with the previous point. Obviously if a deepfake of a person is generated without that person’s agreement, it can lead to discreditation and literally ruin someone’s reputation. Other than that it increases the cases of fraud as falsifying someone’s identity actions without that personas approval might be taken.
What can we do about it?
All that being said, the future may look like a digital nightmare and yes, it is true that with evolving technology it is already quite impossible to detect the difference between false and true. That’s why we have to adapt and take measures, to reconstruct trust towards the media and implement a new structure. That includes honesty from the side of the media, indicating transparently when images are true and when they are manipulated with artificial intelligence.
On the other side people have to be educated and get used to always contrasting information using different sources to be able to get a broader vision of a topic. Of course that doesn’t prevent deepfakes being used to create damage, but it increases the sensitivity about this topic.
Also we have to make artificial intelligence our ally and get used to verify information using tools like WeVerifys Deepfake Detector, Deepware or Microsoft Video AI Authenticator. So companies should implement policies against desinformation and prepare action steps in order to react as fast as possible once false information on the internet is detected in order to eliminate it as soon as possible.
As we mentioned before, any technological progress can be used in different ways, so if we want to make use of the benefits that artificial intelligence and the creation of deepfakes bring us, we also have to find a way to deal with the negative consequences it may have.
