Today we're looking at the paper

"Model Misalignment and Language Change: Traces of AI-Associated Language in Unscripted Spoken English"

by Bryce Anderson, Riley Galpin, Tom S. Juzek

https://arxiv.org/abs/2508.00238

This academic paper explores whether Large Language Models (LLMs) like ChatGPT are influencing human language use, specifically examining if a distinctive "AI-associated" vocabulary is seeping into unscripted spoken English. Researchers analysed over 22 million words from conversational science and technology podcasts, comparing lexical trends before and after ChatGPT's 2022 release. Their findings indicate a moderate but significant increase in the use of words commonly found in LLM outputs, suggesting a convergence between human and AI word choices, unlike baseline synonyms which showed no such directional shift. The study acknowledges the short timeframe and the challenge of definitively attributing causation, highlighting that such changes might represent either a novel AI-driven linguistic shift or an acceleration of natural language evolution.

Hosted on Acast. See acast.com/privacy for more information.