The AI program invites listeners into a simulated American county with unique characters and stories.
It’s the night of September 30, 1840. A light-skinned man with long blond hair and brown eyes lies in a makeshift bed in a distant campsite. A female speech synthesizer tells me that this man has no parents nor any friends or enemies. His name is Jonathan Patience and he is a computer-generated fictional character living in Sheldon County number 1,515,459,035. I am listening to this description of Jonathan’s life on a podcast called Sheldon County. Its host and creator is an artificial intelligence named SHELDON.
SHELDON was created by University of California, Santa Cruz PhD student James Ryan as part of his thesis. It sifts through the experiences of characters living within procedurally generated American counties and creates storylines based on their experiences. The characters in these counties have their own lives and make their own decisions. They interact with one another and even possess unique value systems and goals. SHELDON then turns these stories into a narrative driven, Twin Peaks-inspired podcast called Sheldon County.
Eventually, Ryan hopes to create a unique podcast for each listener, so that they will all hear the stories of different Sheldon Counties.
“A major goal of mine is to demonstrate that generative media—with procedural narrative, generated text, and synthesized voices—can actually be compelling,” Ryan told me in an email. “I'd be thrilled if Sheldon County lead other designers to consider integrating more generativity into their own work.”
SHELDON works alongside Ryan’s other program, called Hennepin, which is tasked with creating these simulated counties and their inhabitants. Ryan said he named SHELDON after Sheldon Klein—an early pioneer of expressive artificial intelligence.
Each simulated county features its own AI-generated characters with their own individual stories. When you listen to your first podcast, SHELDON assigns you a county. Each following episode in the series then depicts the lives and actions of that county’s particular inhabitants.
“When SHELDON writes an episode, the particular content selected for that episode may suggest ideas for a new episode,” Ryan told me in an email. “For example, if there is a side character in the current episode, writing about that side character may cause SHELDON to consider writing a future episode that explores the backstory of that side character.”
Ryan has a history of creating software generated storyworlds. In 2014 he joined the Expressive Intelligence Studio at UC Santa Cruz, which uses artificial intelligence technologies in expressive media. Ryan said that Sheldon Country builds off his previous works, World and Talk of the Town. World was a simulation that followed the lives of tiny abstract agents on an archipelago island. Talk of the Town evolved from this and replaced the archipelago setting with American counties.
On February 22, Ryan released a proof of concept pilot version of Sheldon County onto Soundcloud.
“This is your very own Sheldon County,” the synthesized voice on the podcast told me. “It is yours and yours alone. It is a universe that is accessible only to you. No one has beheld it before and no one will behold it again.”
While the pilot version on Soundcloud is the same for everyone, Ryan told me that he expects his final project to be unique to each listener. Ryan says that listeners will be randomly assigned to a particular county. Once the listener is assigned a county all of the audio files for that particular county will be generated and then uploaded to an RSS dedicated specifically to that county.
Ideally, Ryan said that an email would be sent to listeners with instructions on how to find their series on a popular podcasting platform like the iTunes Store.
Ryan told me how each listener's experience can vary significantly as a result of feedback mechanisms driven by character actions.
“The characters may decide to start a sort of utopian town that is built on the ideals they hold in common,” Ryan said. “So, for example, there could be a group of characters who all believe in law and order and stoicism, and who all despise merriment, so they start a town rooted in these principles. Critically, characters in the simulation hold town hall meetings, where they may propose new legislation and vote on it; if the legislation passes, it can actually affect how life in that town is simulated.”
Ryan let me listen to the follow up to the pilot episode called, ”The good stuff.” From the first few moments the mood and atmosphere was completely different. While the first episode treaded lightly with a sense of dreamlike wonder the second release began with dark foreboding music and a distorted child’s voice. The narration, which was both parts poetic and cryptic, followed the life of Sheldon County resident Charlie Dobes.
Knowing that this entire creation was automatically generated by software makes the experience all the more creepy.
Ryan will complete his thesis this summer and is looking to release a beta version of the podcast for listeners in early 2019. Until then, Ryan said he will continue releasing example podcasts on SoundCloud and Twitter.