conv.

All stories
AIQuiet 7d · day 7

Cornell researchers explain why AI chatbots keep inventing 'Elias Thorne'

A study of 20,000 AI-generated stories reveals how safety training and dataset recycling create a narrow pool of ideas that all LLMs draw from.

newest coverage 7d agosummary written 6d ago3 pieces20 likes & upvotes9 comments MastodonHacker NewsNewswires

What to know

  • Cornell researchers found that major LLMs (OpenAI, Anthropic, Google) repetitively generate the same fictional character—Elias Thorne—in 88% of sampled stories, appearing as lighthouse keeper, clockmaker, librarian, or explorer.
  • The phenomenon stems not from external source material but from safety/copyright filtering that narrows training data, combined with LLMs training on earlier AI-generated datasets, creating a recycled and depleted information pool.
  • The character has escaped chatbots into published AI-slop products on Amazon, YouTube, and health guides, demonstrating how the bottleneck propagates commercial-grade misinformation.

Cornell University researchers Study authors404 Media Investigative reportingDaniel May Software engineer

Cornell researchers explain why AI chatbots keep inventing 'Elias Thorne'
vice.com

How it unfolded 1 development · click the chart to see its coverage articlesposts

Peak 3 pieces in two hours at Sep 19, 2 AM; 3 pieces over 7 days (1 article · 2 posts) Sep 19, 2 AM — 3 pieces · 1 article · 2 posts — Hacker News 1, Mastodon 1, Newswires 1Sep 19, 4 AM — quietSep 19, 6 AM — quietSep 19, 8 AM — quietSep 19, 10 AM — quietSep 19, 12 PM — quietSep 19, 2 PM — quietSep 19, 4 PM — quietSep 19, 6 PM — quietSep 19, 8 PM — quietSep 19, 10 PM — quietSep 20, 12 AM — quietSep 20, 2 AM — quietSep 20, 4 AM — quietSep 20, 6 AM — quietSep 20, 8 AM — quietSep 20, 10 AM — quietSep 20, 12 PM — quietSep 20, 2 PM — quietSep 20, 4 PM — quietSep 20, 6 PM — quietSep 20, 8 PM — quietSep 20, 10 PM — quietSep 21, 12 AM — quietSep 21, 2 AM — quietSep 21, 4 AM — quietSep 21, 6 AM — quietSep 21, 8 AM — quietSep 21, 10 AM — quietSep 21, 12 PM — quietSep 21, 2 PM — quietSep 21, 4 PM — quietSep 21, 6 PM — quietSep 21, 8 PM — quietSep 21, 10 PM — quietSep 22, 12 AM — quietSep 22, 2 AM — quietSep 22, 4 AM — quietSep 22, 6 AM — quietSep 22, 8 AM — quietSep 22, 10 AM — quietSep 22, 12 PM — quietSep 22, 2 PM — quietSep 22, 4 PM — quietSep 22, 6 PM — quietSep 22, 8 PM — quietSep 22, 10 PM — quietSep 23, 12 AM — quietSep 23, 2 AM — quietSep 23, 4 AM — quietSep 23, 6 AM — quietSep 23, 8 AM — quietSep 23, 10 AM — quietSep 23, 12 PM — quietSep 23, 2 PM — quietSep 23, 4 PM — quietSep 23, 6 PM — quietSep 23, 8 PM — quietSep 23, 10 PM — quietSep 24, 12 AM — quietSep 24, 2 AM — quietSep 24, 4 AM — quietSep 24, 6 AM — quietSep 24, 8 AM — quietSep 24, 10 AM — quietSep 24, 12 PM — quietSep 24, 2 PM — quietSep 24, 4 PM — quietSep 24, 6 PM — quietSep 24, 8 PM — quietSep 24, 10 PM — quietYesterday, 12 AM — quietYesterday, 2 AM — quietYesterday, 4 AM — quietYesterday, 6 AM — quietYesterday, 8 AM — quietYesterday, 10 AM — quietYesterday, 12 PM — quietYesterday, 2 PM — quietYesterday, 4 PM — quietYesterday, 6 PM — quietYesterday, 8 PM — quietYesterday, 10 PM — quietToday, 12 AM — quietToday, 2 AM — quietToday, 4 AM — quietToday, 6 AM — quietToday, 8 AM — quietToday, 10 AM — quiet 1
Sep 20Sep 21Sep 22Sep 23Sep 24yesterdaynow · 12:37 PM ET
  1. background

    Elias Thorne spreads beyond chatbots into published AI-generated products — 404 Media documented the character spreading across Amazon books, YouTube videos, and health guides. Software engineer Daniel May identified the name appearing in questionable health content, showing how the recycled character breaks containment into commercial AI-generated slop.

  2. background

    Researchers trace the pattern to AI safety training and dataset recycling — Rather than learning 'Elias Thorne' from existing internet sources, researchers theorize the phenomenon results from safety and alignment training that restricts models away from copyrighted material and adult content, creating a shallower pool of training data. Additionally, modern LLMs trained on datasets built from earlier AI systems amplify this bottleneck by recycling the same limited ideas repeatedly without diversification.

  3. 1

    Cornell study identifies 'Elias Thorne' phenomenon in AI-generated stories

    Researchers at Cornell University examined roughly 20,000 AI-generated stories from major LLM providers and found that specific names and occupations appeared with striking consistency across all models. The name 'Elias' combined with occupations like 'lighthouse keeper,' 'clockmaker,' and 'librarian' appeared in 88 percent of stories, with Elias the lighthouse keeper appearing in nearly two-thirds.

    “Elias Thorne is a weird quirk of the system, but also a symbol of just how hollow and deeply unoriginal chatbots can be.”
    — VICE