Artificial intelligence has fascinated researchers and developers for decades. Currently, algorithms have been developed that allow our phones to advise us in real time on the ideal exposure for a photograph, or even the implementation of driving assistants. However, one of the great obstacles to the development of these calculations by an AI was that of the ability to imagine. A handicap that seems to be about to break through the use of a technique known as "antagonistic generative networks" or GAN. Creating something out of nowhere seemed the last frontier for AI, until now. By confronting two neural networks, the system is able to create increasingly realistic images that deceive the other network. Companies like Nvidia have been able to teach one of these GANs to generate hundreds of faces of people that don't exist. The applications that such an intelligent AI, capable of creating on its own without the need for human help, are spectacular, for example for the world of video games.
5 years ago by enit (39)