Chatbots Keep Telling Stories About Lighthouse Keeper 'Elias Thorne'. We Might Know Why
ChatGPT, Gemini, and Claude are all telling the same made-up story. Here's what it reveals about how LLMs actually work.
Why it matters
A widespread phenomenon where multiple LLMs independently generate identical fictional narratives reveals critical insights into training data contamination, memorization patterns, and potential systemic biases in how foundation models learn and reproduce information.
The key facts
10 to knowCharacter 'Elias Thorne' appears across ChatGPT, Gemini, and Claude independently
Lighthouse keeper and clockmaker narratives are recurring across multiple LLMs
Fiction from chatbot outputs has migrated to real-world publishing (Amazon books)
Researchers investigating root cause of synchronized hallucination pattern
Suggests training data overlap or common contamination source across models
Multiple LLMs (ChatGPT, Gemini, Claude) independently generating identical fictional narrative about 'Elias Thorne' character
Fictional character migrated from chatbot outputs to real-world publishing (Amazon books)
Researchers actively investigating root cause of convergent hallucination behavior
Indicates potential training data contamination or emergent cross-model pattern replication
Raises questions about LLM interpretability and how models encode/reproduce learned patterns
Go to the source
404 Media404media.co
Publisher excerpt: LLMs including ChatGPT, Gemini and Claude are obsessed with telling stories about lighthouse keepers and clockmakers, and one character named 'Elias Thorne' has made his way from chatbots to Amazon books. Researchers are trying to discover why.
