The Language of AI Could Change How Humans Speak (www.schneier.com)

Recent discussions highlight the significant influence that large language models (LLMs) like ChatGPT could have on human communication. Since these models are primarily trained on written text—from books to social media—they capture only a narrow slice of human language, largely missing the unscripted nuances of everyday conversation. As LLMs become more integrated into our communication processes, there is a risk that they may shape our linguistic habits, leading to an erosion of conversational richness, increased curt interactions, and a shift towards more authoritative speech patterns reminiscent of AI directives. This shift is concerning for the AI/ML community, as it underscores how reliance on LLMs may foster confirmation bias, normalize aggressive or curt communication, and diminish critical thinking. The phenomenon of automated jocular or authoritative responses could warp our interaction styles, leading to less empathetic discourse. As AI-generated language increasingly influences our daily expressions, understanding and addressing these effects becomes crucial for ensuring that technology supports rather than undermines authentic human communication, highlighting the need for innovative solutions to train AI on more representative samples of natural speech.