Appropriately on Halloween, researchers from the Massachusetts Institute of Technology have taught Artificial Intelligence to write scary stories. Every hour, a new story is posted on Twitter. Some of them are really scary.
The question that faced the developers of the MIT Media Lab was whether they could teach a machine to scare people. For evoking emotions such as fear and fear would be an art previously reserved for human creativity alone.
The expresses itself still above all in horror and horror stories. Not only horror icons like Stephen King, HP Lovecraft or Anne Rice prove that, but countless hobbyists who share their work in communities like the Reddit group r/nosleep. Every day new and always surprisingly good stories are published there. The latter gave the developers of artificial intelligence to read.
Through a deep learning process, the artificial mind should take some inspiration with it. For this, the artificial intelligence determined among other things frequently occurring phrases. It captured closely word pairs, combinations of specific subjects, or even effective opening and closing sentences. This should help her to understand what causes tension and horror. Fittingly, the developers christened their artificial intelligence Shelley – free after Frankenstein author Mary Shelley, who would have turned 220 this year.
Since October 23, Shelley posts a new story every hour on Twitter, each containing several individual tweets. Often consisting of more randomly mixed word pairs and sentence fragments. For example, report of a woman in a black dress, dull razor blades or a box full of stuffed animals. This makes most of the stories more disturbing, silly and surreal than really scary. But some of the short stories also surprise and could almost come from a human author.
“I heard my mother’s voice scream from the darkness. , God dad, please do not hurt me again. Please do not leave me alone. Please. I do not want to be here anymore. ‘ I turned to my left, and for a second split, I could see it here. After all that time, I could make those black eyes above me.
My heart is beating so fast it is a bit shorter than my breathing. I think I’m being stalked. I do not know what happened to me, or why I was scared of it, but I need to know why. I need to know what happened.”
According to MIT Media Lab developers, readers can also help Shelley become a better author. Namely, by responding to their tweets or tapping their own stories under the hashtag #yourturn to complete Shelly.
They are then also summarized on a separate page highlighting the sentences of the human authors and Shelley contributions. Thus, the Twitter users also provide examples where the AI writer or their elements can take them in future stories.