A new website that uses artificial intelligence can endlessly generate the faces of people who do not exist.
The site, simply titled thispersondoesnotexist.com, is possible thanks to the machine learning algorithms recently published by the Nvidia technology company.
Artificial intelligence analyzes countless photos of the human face to generate realistic images. While the creation of such images initially required advanced computer hardware and specific knowledge, the process is now widely available through the site.
Philip Wang, software engineer at Uber and creator of the website, told Motherboard that the new service is designed to "create a random face every two seconds."
"Most people do not understand how much AI will be able to synthesize images in the future," said Wang.
Concerns have already been expressed about the potential use of this technology by trolls and sock puppets seeking to hide their true identity online.
– x0rz (@ x0rz) February 15, 2019
Others have however noted that the algorithm, although impressive, was still far from perfect.
Aric Toler, researcher at Bellingcat, was able to identify several anomalies in some of the generated photos.
This generator of random faces could make life harder for computer scientists, but there are some dead giveaways with some of these fake faces. Examine the ears, teeth and neck, especially for some monster entrances.https: //t.co/qV2r2ron18 pic.twitter.com/6JVopjygly
– Aric Toler (@AricToler) February 15, 2019
Wang, who claims that artificial intelligence can also be used to create fake "cats, cars and bedrooms," remains for the moment with human faces.
"Faces are the most important elements of our cognition, so I decided to set up this specific pre-formed model," Wang said.
The techniques used are similar to those used when creating deepfakes, videos in which a person's face can be superimposed on another's body.
Technologies have even been developed to simulate a person's voice, giving the impression of saying something they have never really done.
H / T motherboard