Deepfakes, credibility at risk
Everyone at this point has seen the famous app called “Faceswap” where two persons exchange their faces. As a cool party trick is a funny one. But the technique and applications of this replacement of a face for another one is a controversial one.
A similar concept is used in the so called “Deepfakes”. Another application that can exchange the face of an existing video and replace it with another one. At first, it was very buggy and you could detect that the image or video was altered. But with the development of the processing of information technologies and the improvement of the program lead to the point where you can’t distinguish that easily if the video is real or not.
Deepfake is a technique that involves artificial intelligence to create human image synthesis in order to make it believable at the point of not being able to recognize that the images are not real. The trick resides in the machine learning technique called “generative adversarial network” (GAN) which needs existing images or videos to superimpose them into another image or video. The result being a video depicting someone saying or doing something that never happened in reality.
The most popular use for this, is to create pornographic videos of a person performing sexual acts, next to alter words and gestures of a public speak of a politician or public figure. Because the algorithm needs a lot of pictures of someone in a variety of lighting, the celebrities are the main focus of the creation of this kind of videos, because of the vast library that the internet is.
In January 2018, FakeApp was launched. With TensorFlow of Google, used to program the famous DeepDream. In order to work, the program needs a lot of visual references from video or photos to learn both the original and the replacement faces.
This app, opened the ability to create deepfakes to the public in a medium difficulty. Opensource alternatives of this are DeepFaceLab, Faceswap and myFakeApp.
The increasing rate of growing of the technologies and computer applications will bring upon us powerful computer processor and polished apps. So, the deepfakes will be eventually, be completely indistinguishable from real videos. In that case, i think it is needed to educate the next generations in critical thinking and to check the sources of any kind of video in order to verify if such thing is real or not.
References:
https://en.wikipedia.org/wiki/Deepfake
https://www.bbc.com/mundo/noticias-43975322
