The fake mind has been taken on an endless road from the fictional to covering a wide range of categories from logistics, mental data analysis, chess or disease diagnosis in order to name a number. Image generation has been one of the areas of study of several firms, and Nvidia AI, similarly, reached a furor in the development of frighteningly close to the reality of human faces.

Tero Carras, Samuli Laine and Timo Aila from Nvidia have proven the progress made by their AI in the latest technical documentation. The results really are shocking, considering that precedent, the fact that AI made the faces of the people that are not really there. Nvidia studies were built on the basis of the work of Ian Goodfellow in 2014, introducing the basics of transferring manners in a modified version of the generative adversarial network, still familiar as GAN. While GAN generates images subsequently viewing the source, a network of Discriminators tries to separate the fake images from the near-reality ones.

To This NVIDIA scientists have added the basics of transfer manners, in order to combine the properties of the 1st image to another. In fact, 2 neural networks in combination with the transmission of manners have all chances to generate quite close to reality images of people, of which there are none. Nvidia scientists have tried to do the same with images of cats, but cats have proved difficult so far.

AI is very far away?
Fake announcements have been an increasing challenge in the direction of the last few years. Deepfakes is harder to identify, and AI Nvidia who is able to generate images of non-existent people, it has the opportunity to present huge challenges in the future. There have been a number of presentations of researchers capable of making fake videos with the role of well-known politicians, and if this development is used, it is possible to apply it to the creation of fake videos that look even more close to reality.

Not least, the cat world will likely not need to worry about Nvidia's AI. For now.