Please note: This story links to the Nightmare Machine website, which includes a warning stating that some photos may contain scary content and may not be suitable for all users.
When it comes to holidays, MIT is known more for celebrations of its own creation, such as Pi Day and the Baker House Piano Drop. But a trio of Media Lab researchers are adding an Institute flair to this year’s Halloween season—using a fright-soliciting algorithm called the Nightmare Machine.
The Nightmare Machine uses a deep-learning algorithm created by Associate Professor Iyad Rahwan and post-doctoral researchers Pinar Yanardag and Manuel Cebrian to detect our deepest visual fears. According to its website, the Nightmare Machine takes ordinary photos of people and landmarks and uses artificial intelligence to determine how their photos would look if the people and places were “haunted.” Users can visit a dedicated voting page to help the algorithm learn which images are the scariest.
“Clinton, Drumpf, the White House too, terrifyingly transformed by MIT’s ‘Nightmare Machine,’” Washington Post, Oct. 24, 2016
We use state-of-the-art deep learning algorithms to learn what haunted houses, ghost towns or toxic cities look like, Pinar Yanardag said.
The algorithm extracts elements — such as a bruised-black palette — from these scary templates and implants them in the landmarks.
The Machine’s Instagram page contains more than 100 so-called “nightmarified” images, including a darkened version of MIT’s Great Dome shadowed by an ominous black sky and a crumpling and desolate MIT Media Lab. The page is heavy on non-MIT imagery, including a distorted Hillary Clinton and Donald Trump debating over a floor of skulls.
According to the Nightmare Machine website, the computer generated images are powered by deep learning algorithms with a touch of evil spirits. The website also includes a 2,000-year timeline that merges the origins of artificial intelligence and Halloween.