For years, artificial intelligence has been increasingly influencing our everyday life, but also the economy, politics and science – as chatbots in customer service, GPS assistants or characters in video games.
AI can be used in many different ways and offers us new innovative solutions to problems in the most diverse areas. However, deepfakes show that AI also offers potential for controversy.
Deepfakes: Huge Potential In Both Directions
The term »deepfake« consists of the words »deep learning« and »fake«. For a deepfake, media content is manipulated using artificial neural networks. There are three different types of deepfakes.
Face swapping means replacing the face with another generated face. The generated face can then take over the movements and reactions of any person or of the manipulated person which makes the videos look deceptively real although they are not.
Besides face swapping, there is also voice swapping. This involves voices being changed or artificially simulated. Furthermore, so-called body puppetry allows to transfer body movements to other people.
Creating a deepfake is relatively simple. Over the last two years, some apps have emerged that enable users to create deepfakes without any prior knowledge. Moreover, AI frameworks such as Google’s TensorFlow, which allow for the creation of deepfakes, are easily accessible and free of charge.
Deepfakes: Huge Potential, Huge Risk
The deepfake phenomenon has not existed for a long time. The term has only been circulating on portals such as a YouTube, Twitter or Reddit since 2017. A huge part of deepfakes is related to pornography. The faces of popular pornstars are replaced with the faces of famous actors/actresses.
Deepfakes also carry huge risks for politics and the economy. Due to the difficult detection in combination with rapidly spreading news over the Internet, it is not unlikely that some fake news briefly cause a stir because they seem to be real. In this video, the YouTuber Ctrl Shift Face shows how real deepfakes can actually be:
Despite the dangers of deepfakes, they have huge potential for science and the economy because not only people but also landscapes can be manipulated with this technology.
Landscapes can thus be virtually shifted into all types of weather conditions. This allows for even more versatile training, in particular in industries that depend on the training of their AI, because in addition to tests in the real world, AI is also capable of learning via simulations.
The deepfake phenomenon is interesting and shows how versatile and powerful artificial intelligence already is these days. In the future, however, measures will have to be taken to be able to easily identify deepfakes.
Our 5 Reading Tips Of The Week
What are deepfakes – and how can you spot them? [The Guardian]
So echt sehen Deepfake-Videos aus [Handelsblatt]
Ctrl Shift Face [YouTube]
Auf dem Weg in eine alternative Realität? [Tagesschau]