Deepfakes and the intensification of fake news

Recently, a number of fake celebrity videos in which they are portrayed doing things or defending opinions they disagree with have come online. These videos, which, apart from their appearance, have nothing to do with  the personalities portrayed in them, were called deepfakes, and represent a major concern about individual image rights today, among other potentially dangerous repercussions. This post aims to present information about the technology used in deepfakes, as well as to develop the theme of the possible political and social repercussions of this technology – especially in the context of the intensification of the use of fake news for social and political manipulation purposes.

What are deepfakes?

Deepfakes are, basically, videos  they are digitally manipulated to make people seem to say or do things that aren’t really real. In other words, they concern a technique of very advanced visual effects, which allows to juxtapose a person’s face to the body of an actor that was actually filmed during the production of the original video.

These effects are achieved through the use of artificial intelligence through machine learning techniques. What happens is that the software used recognizes the face of the person who acted in the video and applies a 3D model of the face of the person to be inserted in the file. This 3D model, in turn, is created by analyzing a vast database containing several photos of the person who will be introduced to the video.

As the quality of deepfake is directly related to how much the machine learning algorithm can learn about the analyzed face, the availability of a very significant amount of photographic references of the person is necessary to obtain a realistic representation. This means that, in general terms, the main people affected by the technology are the public people, often photographed and widely publicized – these correspond to the richest demographic in reference to creating compelling deepfakes.

Image rights concerns

Deepfake technology has several benefits when used with the consent of the person to whom it will be employed. Many of them point to significant advances in the field of film production, with the possibility, for example, of altering the movement of actors’ lips to match what is being spoken in versions dubbed in other languages ​​- eliminating the lip dyssynchrony that is usually inevitable in this kind of conversion.

Another possibility from this technology is the general improvement in the quality of human faces that we will be able to generate through the use of computer graphics, which in itself represents an invaluable advance in the creative possibilities for digital media. This, in turn, represents a major step towards democratizing access to these visual possibilities – which were previously restricted to large studios, holding unimaginable amounts of financial resources, but can now be made accessible to smaller and even independent producers in a much more affordable way.

Finally, there is also the possibility that the benefits of this technology extend far beyond the entertainment industry. If the technology evolves enough to be used in real time, it can be used for videoconferencing therapy sessions for people who are uncomfortable exposing their faces, for example, or even conducting more impartial  job interviews, without gender or race , etc.

On the other hand, an important concern that arises with the popularization of such a technology is precisely the misuse of people’s faces for the production of content with which their respective owners did not consent. Returning to the topic of deepfakes involving celebrities and public people such as politicians, prominent business owners, etc., it is easy to imagine how this technology can generate conflicts over image rights. This is because the widespread availability of photographic material from these famous people can subsidize the production of videos that not only violate image rights but also can convey slanderous, fake content and in many other ways designed to mislead the audience and cause some kind strategically conditioned reaction. Which brings us to the next topic …

Fake news 2.0

The fake news phenomenon is already widely known to most people. Usually, however, such news spread over the Internet as a means of generating misinformation and manipulating masses of people most susceptible to their nuances were made available in the form of images. That’s because video represents a much harder media to manipulate, requiring much more technical knowledge from the creator to produce something minimally believable.

In this sense, videos have always been endowed with a certain presumption of truth – often being presented as thorough proof that some information is true. In the context of deepfakes, however, this logic falls apart, and what we have left is fake news with devastating potential, surrounded by this aura of veracity, precisely because of the format of the media.

It’s important to remember that fake news has already been responsible for social manipulation to levels never before imagined. This concern, in general terms, gained strength with the Cambridge Analytica scandal, in which this company, which used Facebooks ecosystem and user data, managed to manipulate the results of extremely important political votings – notably the 2016 US presidential elections and the Brexit voting, both known to be affected by Company activities.

In the current context of the US electoral race, therefore, deepfakes are an imminent danger, but this situation is not limited to foreign territory. In Brazil, the growing political-ideological polarization that we have experienced in recent years is fertile ground for the proliferation of false news of all kinds – and deepfakes pose a problem in that they can manipulate the population even more efficiently than regular fake news, potentially amplifying this polarization of political groups in the country.

In this sense, it is important that we increasingly seek to confirm the veracity of the content available to us online. There are, for example, fake news verification portals such as FactCheck.org or Fact Checker, among others. In addition, there are specific features that we can look for in videos to try to find out if the face we see was superimposed on one that was originally there – the eyes being the biggest evidence that the video was manipulated (usually deepfakes may seem to be looking different directions from those that the face of the individual points, and may blink in unnatural patterns).

Conclusion

In summary, it can be said that now that it has been made available for general use, deepfake technology is a reality that we will always have to deal with. Knowing that, it is important for us to be aware of its existence, and to know that, contrary to what was thought some time ago, there is no totally credible information on the internet, and even videos may have been manipulated to convey misinformation.

It is therefore up to us to take appropriate precautions to keep ourselves informed and to avoid sharing false content, as well as to point out when something of this nature is shared by acquaintances and even report the posting when it is clearly malicious. It is true that deepfakes are a potential danger, but the habit that many of us have already established to verify the truth of information may be enough to prevent technology from being abused for illegitimate purposes.

Did you like this post? Are you interested in topics related to artificial intelligence? Click here for more IRIS content on this subject!

The views and opinions expressed in this article are those of the authors.
Illustration by Freepik