FEED Issue 13

Words by Neal Romanek

BREAKING NEWS FROM THE STREAMING SECTOR

THESE PEOPLE DO NOT EXIST

With every hit of the refresh button, thispersondoesnotexist.com produces photorealistic human faces from scratch. Though the faces appear to be a set of randomly selected headshots, they are actually the product of an AI algorithm called a generative adversarial network (GAN). GANs were first introduced as a concept in 2014 by computer scientist Ian Goodfellow. They consist of two neural networks: the generator and the discriminator. These computer

programmes compete against each other millions upon millions of times to refine their image-generating skills until they’re able to create fully fledged human faces. Since its conception, Nvidia – designer of GPUs for the gaming market – has been at the forefront of this technology. Researchers for the company have led multiple GAN studies but, until recently, weren’t able to create high-quality, 1024x1024 images using this method. In late 2017, Nvidia cracked the

code using a technique called StyleGAN, which builds on the original technique by giving researchers more control over specific visual features. The power of algorithms like StyleGAN has raised a lot of questions. While it has the potential to increase artistry and revolutionise video games and virtual worlds, it could also be used for more dubious purposes, like political propaganda and influencing online campaigns. Previously, GANs have even been used to create deep fakes.

feedzine feed.zine feedmagazine.tv

Powered by