The curious case of Elias Thorne – and what he tells us about AI inbreeding | Arwa Mahdawi
TL;DR
Arwa Mahdawi’s Guardian column turns Elias Thorne into a symbol of a strange sameness in AI writing: when chatbots get simple story prompts, the character keeps appearing as a lighthouse keeper, clockmaker, detective, or similarly familiar figure. A Cornell study sampled 20,000 stories from four current models and found that the name Elias appeared in 26.5% of them. 88.3% contained at least one of 11 core words, including Elias, Mara, Elara, lighthouse, keeper, and clockmaker.
Nauti's Take
The interesting part is not whether Elias Thorne is mysterious. The interesting part is how quickly a harmless pattern becomes the default imagination of multiple systems.
This is exactly the kind of AI slop that looks fine at first and then starts smelling the same everywhere. Anyone using AI for writing needs sharper prompts, original source material, human editing, and the discipline to remove generic smoothness.
Briefingshow
Elias Thorne is more than a funny internet pattern. The case shows that AI models do not simply mirror the full breadth of their training data; alignment, preference data, and model-to-model imitation can steer them into narrow output lanes. For media, education, and content production, more AI can end up meaning less variety.
Sources