Editor's note: This story was originally published Nov. 8, 2017. We're bringing it back today to offer you some Halloween heeby-jeebies.
WASHINGTON (CIRCA) - Shelley isn't your average person.
Actually, she's not a human at all. She's a robot that uses artificial intelligence to write and share hair-raising horror stories. Named after famed "Frankenstein" writer Mary Shelley, the bot is the brainchild of Massachusetts Institute of Technology postdoctoral candidate Pinar Yanardag and research scientist Manuel Cebrian. They claim that Shelley is the world's first collaborative AI horror writer.
"The idea of launching Shelley was doing follow-up work of what we did [in 2016] -- the Nightmare Machine project -- where we created scary images using AI. And, [in 2017], we wanted to do a text version where we create horror stories using AI."
On Oct. 27, 2017, Shelley came to life. But before she did, co-creators Yanardag and Cebrian spent a lot of time figuring out where to obtain enough data to power Shelley. The more data they fed into the machine, they explained, the more it would "learn" and mimic. That's why they eventually decided to pull data from the popular Reddit forum "No Sleep." Throughout the past 10 years, users have been sharing and collaborating on their own horror stories. As a result, Yanardag and Cebrian were able to feed Shelley nearly 700 megabytes worth of data. That's equivalent to a feature-length film.
After combing through the data, Shelley was able to pick up on certain patterns used in the Reddit stories, which helped her develop her own way of processing information. For instance, those algorithms helped her develop a sense of linguistic intelligence -- allowing her to determine the next likely character.
"So, you can basically think [of it] as a computer program that is able to process the data that we feed to it," Yanardag said. "And then it starts from a random character and it creates where the character would go next and next."
Upon first reading Shelley's dark and twisted stories, some may find the robot's coherent verbiage surprising. Her stories, which are gathered online, read as if any other human created them. Yanardag explained that's the result of the amount of data used to train Shelley.
Shelley, like many humans, spends most of her time on social media -- on Twitter, in particular. Her co-creators designed her so that she would be able to tweet and collaborate on sick-twisted tales with real humans. About every hour, Shelley would tweet the beginning of a story and invite others to finish the story by using the hashtag #YourTurn.
Roughly one week after her introduction, Shelley had generated nearly 1,300 tweets and attracted 6,000 followers. Her co-creators handpicked some of their favorite stories and shared them on Instagram.
Shelley's seemingly infinite creativity offers her co-creators a unique opportunity to further analyze and develop artificial intelligence.
Even though Shelley hasn't tweeted since Nov. 12, 2017, the story has just begun for her co-creators.
"We need to actually analyze the data clinically. We really need to understand how Shelley works, because even though we created her, we don't really understand how's she able to produce this, so we need to do psychological investigations. We must work on understanding how she works, and how she works with people [in the] next [few] months."