Caine Prize AI-writing claim didn't rest on Pangram alone: stylistic and stereotype evidence cited
TuhinChakr · x · 2026-11-18
Will Oremus clears up a misreading of Vauhini Vara's Atlantic piece on the Caine Prize winner: critics assumed she relied only on the Pangram detector to flag the story as AI-written. She didn't — and the screenshot shows her actual textual evidence: the story is full of AI-typical 'not x, but y' constructions, metaphor pileups and platitudinous aphorisms, unlike the author's earlier, plainer prose. More disturbingly, it reproduces outsider stereotypes about African life — the blundering unemployed father, the resilient matriarch dispensing clichéd wisdom — echoing research showing AI products reproduce cultural tropes.
More from Fun
- Merriam-Webster adds 'slopaganda': propaganda made with AI — mattwbaker · 2026-11-18
- spawn 2027 player reports lag after simulating all of NYC with 'biologically accurate' humans on a phone — TAbrodi · 2026-11-18
- Claude Code ported Android 9/10 to an abandoned 2016 tablet with no source code — HomiRuimm · 2026-11-18
- Founder builds a DIY device that lets you scratch Spotify music live — jnack · 2026-11-18
- Tiny 47M-param diffusion model paints pixel-art emoji in your browser — victormustar · 2026-11-18
- Is the 'AI CEO Thanksgiving curse' actually a thing? — nmasc_ · 2026-11-18