Pangram v4 flags pre-2020 thousand-line texts as 100% AI-generated
birchlse · x · 2026-09-23
Antoine Tilloy points out that several thousand-line texts written before 2020 — long before modern LLMs existed — are flagged as "100% AI" by AI detector Pangram v4. The false positives underscore how unreliable such detectors are for long texts and the risk of false accusations.
More from Fun
- Marc Lou's 'A/B testing' advice mocked: ship without tests and still win — jdluk87 · 2026-09-23
- After Opus's music video went viral, a Redditor made one for 'Caveman Twerk' with Astra — eposnix · 2026-09-23
- Restaurant manager on AI booking calls: the fake conversationality feels condescending — annetgriffin · 2026-09-23
- Retro AI Videos From 3 Years Ago Were Somehow Crazier Than Today's — BLUECOW009 · 2026-09-23
- MIT professor jokes his alma mater lost No.1 ranking after 15 years — to his other alma mater — jefrankle · 2026-09-23
- Autonomous Lamp: a $499 open-source companion robot that sees, hears and chats — dee_hw · 2026-09-23