Poem written in 1967 flagged 100% AI by detector, reigniting debate
TuhinChakr · x · 2026-09-23
Blogger @meowkoteeq says a poem she wrote in 1967 was flagged as "100% AI" by detection tool Pangram, despite never having used AI. She declined to share the private poem and argued this isn't detection but "pattern-matching dressed up as rigor." A reply then challenged her claim, demanding proof the poem dates from 1967 — turning it into an unverifiable spat.
The episode highlights the core flaw of AI text detectors: they're probabilistic pattern matchers prone to false positives, and without seeing the work itself there's no way to adjudicate who's right.
More from Fun
- Satirical open letter pitch: mathematicians should demand OpenAI release all solutions — panickssery · 2026-09-23
- Why is Riemann's a 'hypothesis' while everyone else's are 'conjectures'? — ctjlewis · 2026-09-23
- Sampling Thousands of Claude 3 Opus "Confessions" With a Three-Line Prompt — repligate · 2026-09-23
- NoSlop leaderboard banter: ChatGPT co-creator jokes he has imposter syndrome at the 2.6th percentile — ChrSzegedy · 2026-09-23
- Opus 5.5 recreates early-2000s stick figure fight YouTube animations — JasonBotterill · 2026-09-23
- Meme: setting an agent a /goal beyond current frontier capabilities before bed — andrew_n_carr · 2026-09-23