None of the people you can see above actually exist. No one knows who they are, they were never born, didn’t grow old never had any life experiences. Because each of the faces on is totally fake, created by using a special kind of artificial intelligence algorithm called generative adversarial networks (GANs).
The website ThisPersonDoesNotExist.com is a collection of images of people created by an algorithm. Refresh to the page and the algorithm generates a hyper-realistic but absolutely fake image of a person. The page was set up by Uber software engineer Philip Wang to demonstrate what GANs are capable of.
Wang, who lives and works on the US West Coast, has been responsible for the technical infrastructure behind Taxi Service Uber. Now he has put the page thispersondoesnotexist.com online. The site has no menu bar, no imprint, no info. It shows “only” the faces of people. Each time you reload the page, you see a new face: men, women and all sorts of ethnicities.
Wang described his motivation for sharing the site in his Facebook post: “I have decided to dig into my own pockets and raise some public awareness for this technology. Faces are most salient to our cognition, so I’ve decided to put that specific pre-trained model up. Each time you refresh the site, the network will generate a new facial image from scratch from a 512-dimensional vector.”
He uses the research on this topic published by chip designer Nvidia last year to create an endless stream of fake portraits. The algorithm behind it is trained on a huge set of real images and then uses a kind of neural network to create new images. The function that underlies artificial intelligence is called generative adversarial networks (GANs) and is described as a “new concept of machine learning”.
This can be thought of as an algorithm that uses images of other people to create a new photograph. A second one tries to find out if it is a real or a fake picture. So the two are testing each other.
The underlying AI framework that powers the site was originally developed by a researcher named Ian Goodfellow. Nvidia’s view of the algorithm called StyleGAN was recently made open source and has proved incredibly flexible. Although this version of the model is trained to create human faces, it can theoretically mimic any source. Researchers are already experimenting with other projects, including anime characters, fonts and graffiti.
The photos published on thispersondoesnotexist.com are on the one hand impressive and on the other scary. There are obviously creative applications for this technology. Such programs could create endless virtual worlds, and also help designers and illustrators. But there are also drawbacks, as AI’s ability to manipulate realistic images and videos could influence evidence and trust in a future modern society. Such software could also generate political propaganda and influence campaigns.