From the “rise of the machines” department, and coming to a fake news story near you comes this troubling development that blurs the lines of reality and fantasy even further.
Video graphics giant NVIDIA recently released a paper demonstrating a new training methodology for generating unique and realistic looking faces using a generative adversarial network (GAN).
The results are indistinguishable from reality, yet none of these are real people – they are computer generated.
From the paper abstract:
“We describe a new training methodology for generative adversarial networks. The key idea is to grow both the generator and discriminator progressively, starting from low-resolution images, and add new layers that deal with higher resolution details as the training progresses. This greatly stabilizes the training and allows us to produce images of unprecedented quality, e.g., CelebA images at 1024² resolution. We also propose a simple way to increase the variation in generated images, and achieve a record inception score of 8.80 in unsupervised CIFAR10,”
Obviously, this has applications in computer gaming, and when perfected, could replace hundreds of problematic Hollywood actors without risks of the AI generated people enduring waves of sexual harassment put forth by producers to “get the part”. The downside though, is that if people can be this well simulated, this means the value of surveillance video and other court evidence will be worthless in the future. It also means a booming industry in “fake news” video.
Here is a video compiling some of the faces they have generated (best viewed at full-screen, 1080i or better):
Yet as good as these models are, we still can’t get climate modeling right.