Nvidia’s Computer-Generated Faces Are A Creepy Preview Of Technology To Come

Photo: Behrouz Mehri / Contributor (Getty Images)

At the forefront of new graphics technology sits Nvidia, a company that virtually holds a monopoly when it comes to developing hardware that can render realistic 3D environments and models.

While Nvidia’s reputation has mostly been built on its work in gaming graphics where it’s responsible for bringing virtual worlds to life, its products are also used at highly influential graphic design and animation firms (i.e. Pixar). In these instances, technologies such as its progressive growing technique are playing a big part in replacing real people with digital robots created using software and algorithms.

Nvidia first spoke about the generative adversarial networks (GANs) last year when it showed off a series of randomly computer-generated faces that looked almost exactly like what you’d expect on a cover of The National Enquirer. Now, artists are playing around with the publicly available technology with interesting results.

Robot Technology: Experts Say Human-Level A.I. Will Appear In Five To 10 Years

Take, for example, a two-minute video crafted by Twitter user @highqualitysh1t. Using thousands of photos taken at a photo booth in Germany to create a sampling of facial structure data, it shows a stream of animated computer-generated faces that look like they belong in a copy of Scary Stories to Tell in the Dark.

Seen below, the video might not seem consequential, but it’s a warning that in the near future you probably won’t be able to tell whether or not the person you see on-screen is real or an amalgamation of the real and virtual worlds.

Twitter user @Yosun took the creepiness one step further by adding mustaches into the mix:

Another example of the progressive growth of GANs can be seen in the video below. It’s far less bizarre and shows real images manipulated with CGI elements.

Machine learning, a field of artificial intelligence that is a leading effort by mega tech companies like Google, is used to power Nvidia’s iteration of progressive growth of GANs. As described by Nvidia:

The key idea is to grow both the generator and discriminator progressively: starting from a low resolution, we add new layers that model increasingly fine details as training progresses. This both speeds the training up and greatly stabilizes it, allowing us to produce images of unprecedented quality, e.g., CelebA images at 1024²

Dangerous Technology: 6 Robots That Could Destroy You If They Really Wanted To

Similar principles have been used by Nvidia to develop A.I.-powered image analysis that will soon be used in healthcare for high-quality imaging of the heart, liver, and lungs.