[ad_1]
Egor Zakharov created a new artificial intelligence system capable of generating a fake video clip from a single photo. Deepfakes or fake clips that get subjects to say and do things that they do not usually need huge data sets. This is one of the factors that has kept technology safe from bad actors.
RELATED: WATCH SALVADOR DALI COMES BACK TO LIFE BY AI
However, the new software is alarming. Software development by Samsung's artificial intelligence laboratory in Russia has been detailed in a document published on the arxiv website. The author of the article indicates that their research has resulted in a system that first performs a long meta-learning on a large set of video data, and who is then able to frame a learning simple and unique neural talking head models of previously invisible people's accusatory training problems with generators and large capacity discriminators.
Good and bad applications
. able to initialize the parameters of the generator and the discriminator in a specific way to the person, so that the training can be based on some images and be done quickly, in spite of the need to adjust dozens millions of parameters.
We show that such an approach makes it possible to learn very realistic and personalized new head models, even portraits. The software can have fun apps like being portraits to life, as the example given in the video below of the Mona Lisa
. to be used as a tool of political and social manipulation. While most of us are still struggling to understand and monitor the false news in our feeds, deepfakes is at the next level. Some of the examples presented in the video, even from unique images, could mislead many people, especially when they are presented in a context supposed to be trustworthy.
A video was widely shared in Belgium last year in which the President of the Republic, Donald Trump, said: you know, I had to withdraw from the agreement on the climate of Paris ", he said, looking directly into the camera, "and you should do the same. Except that the video was wrong.
It was created by a Belgian political party, Socialistische Partij Anders, or sp.a, and shared through social media. The party had commissioned a local production company to create the deepfake closing software and thought that the poor quality of the clip would be clear to all viewers.
Looking at the clip knowing it's a fake, it's obvious that it's not real, but it seems to have misled many people who expressed their anger that the United States was playing with the climate policy of the European country. History shows that even a mediocre version of deep dummy videos can cause distress. It's unclear what Samsung wants to use the newly created algorithm for, but we can say with certainty that a new era of false information is about to enter.
[ad_2]
Source link
