Pornhub is one of the most visited web pages in the world, with more than 70 million unique users per day, a real nonsense, which has made it the platform par excellence to search for any video, of course sexual content, since any user can make their own creations and upload them to the platform.
But there is a new type of videos, called deepfake, that It is not related with deep throat (Deep Throat) a movie that will probably sound more or less according to your age. Deepfake videos are videos that make use of artificial intelligence to replace the faces of the protagonists with famous actors or actresses.
Video in which we see Jessica Alba's face on the body of a porn actress, using the deepfakes application
Having turned this type of videos into one of the most popular on this platform, and much to its regret (in addition to avoiding future legal problems) Pornhub has announced that it has begun to prohibit all videos that make use of machine learning to exchange the faces of the protagonists for famous actors and actresses, since the company prohibits any type content not consensual by both parties, since logically the owner of the face has not given their consent.
The statement that the company has sent to different media, Pornhub affirms that this type of videos is very similar to the serious problem that affects many users and users called Vengeful porn, a growing problem and that at the moment the large platforms are doing nothing to try to combat it, except the tests that Facebook has begun to carry out through Facebook Messenger in Australia, one of the countries most affected by this problem.
Despite the ban, it will be quite a difficult task for Pornhub to be able to delete all videos of this type, although most of those that are uploaded they are labeled deepfake, so they can be easily found. However, we can also find other videos of this type that use the name of famous actors or actresses to promote their deepfakes, which complicates their elimination unless the platform continuously tracks the name of all actors and actresses, a task that It seems titanic at first.
Origin of deepfakes
Just over a month ago, the website Motherboard discovered a disturbing new trend on Reddit, in which users were creating clips of pornographic videos generated by artificial intelligence. changed people's faces for porn movie stars. But everything changed, when the first deepfake began to circulate in which the star was Gal Gadot, the actress who has been in charge of starring in the movie Wonder Woman.
In this first video that has given rise to this new category within the prone, the face of a pornographic actress had been changed to that of the actress. To create the video were used machine learning algorithms, open source libraries, Google images and YouTube videos.
At first there was only one user who was responsible for creating this type of videos, but today, there are an application that allows you to make this type of video, not only with celebrities but also with ordinary people, an application that according to Motherboard can be used without technical or programming knowledge, which makes its use even more dangerous.
Motherboard has contacted the developer of this application, whose pseudonym is deepfakeapp. He states that he is working to improve the user interface so that creating a video is as simple as selecting the source video, which we want to swap faces, download a specific face from a library from a correlated neural network available publicly and finally swap your face for the famous one with the push of a button.
At first, you may have doubts about the results of these types of videos, but if you do a search you will see how some they give us strangely realistic results. Despite having announced the withdrawal of this type of videos, we can still find videos "starring" Scarlett Johansson, Angelina Jolie, Kate Perry, Natalie Portman ...
Neural networks
In the middle of last year, a research team from the University of Washington created a video making use of artificial intelligencel of the former president of the United States, Barak Obama, making use of the other video that had been previously recorded where the former president appeared.
This research team, it took 14 hours to train a neural networkl so that he was able to synchronize the movements of his mouth and lips with the audio. This system was also capable of adjusting the movements of the head and jaw to offer a more than realistic result.
The result, we can see in the video above, a video that could have been created perfectly making use of the CGI that invades most Hollywood movies, and whose cost is extremely high. But here only a neural network programmed for this purpose was used for a few hours.