Can artificial intelligence write fiction with real tension?

In the first instalment of Synthetic Stories, a new series of audio short stories, horror podcast junkie Amelia is lying in bed listening to a podcast when she begins to hear ghostly voices and haunted house sounds echoing around her room. The following day she tries to put it out of her mind, but, as the voices start up again in the evening, she notices that the AI assistant on her phone has installed an app without her consent called Horror World. Opening it, she finds a portal to a world that appears to be based on her own life, created by a shadowy syndicate of programmers “who had created an AI so advanced that it had gained sentience, and was now using its vast intelligence to manipulate humanity for its own gain.”

Scared? I doubt you will be, though suspense is hard to build in just eight minutes, which is the length of this first story titled “Amelia”. What is scary, however, is the way it has been made. Synthetic Stories is AI-generated, from the plot and the voices to the music and artwork. It arrives at a time of heightened anxiety around AI and its impact on the creative industries. Apple recently launched an audiobook catalogue narrated by synthesised voices, much to the alarm of voice actors noting the threat to their livelihoods. And earlier this year the musician Nick Cave was unimpressed with an attempt to replicate his songwriting style via AI-generated song lyrics, calling AI an “emerging horror” and the song in question “a grotesque mockery of what it is to be human”.

Synthetic Stories is really an experiment conducted by the audio content agency This Is Distorted, to see to what extent the technology can pull the wool over our eyes. The whole series was created in just 24 hours and began with producers asking the AI chatbot ChatGPT to tell them a short story based on the themes of AI and podcasts, and using prompts including apps, horror and dark twists. Later came the music courtesy of AI music generator Soundraw and the narration via the voice cloning app ElevenLabs.

The result is a series that really does sound as if it was made by humans, although that isn’t really an endorsement, since humans have proved themselves more than capable of making terrible podcasts. That the story is told by a single male narrator means the vibe is more audiobook than pod. More pointedly, the voice lacks the intonation and inflection that would be provided by a good voice actor.

As for the story, the pacing is off, it is prone to repetition and the mystery is wrapped up way too soon, though asking an AI bot to create a story about the threat to humanity presented by AI is enjoyably meta. The best that can be said about Synthetic Stories is that it is a worthy experiment that tells us much about the possibilities of AI and, most importantly, its limitations.