Meet the Nightmare Machine, AI’s discovery of our deepest fears

26 Oct 2016

Nightmare. Image: nito/Shutterstock

It was only a matter of time before computers started honing in on humanity’s deepest fears of darkness, fire, war and aliens. Researchers at MIT have gotten the ball rolling.

AI’s latest trick could prove more telling than many before, with MIT’s new Nightmare Machine a perfect marriage between computers and horror.

Essentially a series of algorithms, Nightmare Machine is an attempt to find the scariest faces and locations possible, relying on us humans to make sure it works.

Future Human

Nightmare Machine

Using Google’s Deep Dream approach, and hinting at Microsoft’s ‘How old do I look?’ projects, MIT researchers generate ghastly images of people and distorted, scary portraits of famous locations, putting them to a public vote.

Us schmucks tell the algorithm which images are more scary, thus tailoring it to continue down the path of human suffering.

When given a batch of ten images selected by Nightmare Machine, users rank them. The results inform the algorithm which features are most commonly perceived as scary, allowing it to improve.

Obviously themed for Halloween, it took just three researchers to create this tool: Pinar Yanardag, Manuel Cebrian and Iyad Rahwan.

Relying on deep learning, they taught the algorithm several horror styles for haunted places: alien invasion, fright night, ghost town, haunted house, inferno, slaughterhouse, toxic city and tentacle monster.

paris taj london colosseum

The team charted the history of the horror genre in order to introduce Nightmare Machine into our lives. One of the most notable periods is the winter-long year of 1816, when Mary Shelley, John William Polidori and Lord Byron were locked in a house.

During this time, the trio wrote works of literature that shaped the monster (Frankenstein), vampire and apocalyptic genres for centuries to come.

Byron’s child Ada Lovelace got computer science off the ground soon after, and the rest is Nightmare Machine history.

While the locations initially appear to be the gem in this, it’s the faces that are a real work of computer science art. They look like the work of graphic designers and Photoshop but, beyond the initial algorithmic coding and the introduction of base images, it’s the computer doing all the work.

From these base images, the Nightmare Machine is now advanced in creating its own facial images from scratch.

We are essentially feeding the monster.

Gordon Hunt was a journalist with Silicon Republic