In 2018, A.I. Became Impressively Good at Creating (Creepy) Fine Art

When you think “artificial intelligence” your mind probably conjures images of supercomputers crunching countless arrays of infinite numbers. But in 2018, A.I. made waves in the arts scene with algorithmically generated paintings that are pretty disturbing but can hardly be distinguished from the work of a human.

One specific work titled Portrait of Edmond Belamy sold for $432,500 in October at the Christie’s auction house in New York City. Christie’s called this the first computer-created artwork to be sold by a major auction house, but it’s by no means A.I.’s first foray into the art scene.

Google’s 2015 trippy as hell DeepDream tool was the first time A.I. abstract art made waves online, but the underlying tech was established a year before that. Esteemed computer scientist Ian Goodfellow created the class of A.I. algorithm known as a generative adversarial network or GAN. This framework has been reiterated upon and used to create countless works of art, including the Portrait of Edmond Belamy.

This is #3 on Inverse’s list of the 20 Ways A.I. Became More Human in 2018.

NVIDIA researchers developed their own GAN and published it in April, which enabled A.I. enthusiasts across the internet to feed it their own images to create works of art. Creative developer for the ZKM Center for Art and Media Daniel Heiss plugged in 50,000 photobooth images into the algorithm to create a disturbing but mesmerizing swirling of faces, seen in the video above. He told Inverse that the results were better than he ever expected.

“I saw the crazy warping of one face images into three face images into two face images and so on. That was much better than I ever thought,” he said. “I even tried to filter the images so that only images with one face are used, but while I was working on that the samples generated from the unfiltered dataset came out so good that I stop that.”

While Heiss’ experiment is proof that A.I. has become shockingly good at creating works of art, it’s still up to humans to select what to give algorithms. The future of art could be led by hybrid artist-programmers directing A.I. to create trippy abstract works paintings and portraits.