Academic Research Reveals 48% Surge in Human Adoption of LLM-Specific Vocabulary Post-ChatGPT Release

Michael S. Galpert, founder of an AI product studio, recently made a striking prediction on social media, stating that "> most humans will start speaking like LLMs." This assertion resonates with emerging academic research indicating a measurable influence of Large Language Models (LLMs) on human linguistic patterns, particularly in spoken communication.

Galpert, an entrepreneur with a background in tech and AI, including his current venture Contains Inc. and previous work on Fortnite, frequently shares his perspectives on the evolving impact of artificial intelligence. His tweet highlights a growing anticipation within the technology sector regarding AI's pervasive integration into human interaction and communication.

Empirical evidence supporting this linguistic shift comes from a study by Yakura et al., published on arXiv, which analyzed approximately 280,000 English-language video transcripts from academic institutions. The research identified a significant increase in the usage of words distinctively associated with ChatGPT, such as "delve," "realm," "meticulous," and "adept," following the LLM's release in November 2022. Over an 18-month period, words like 'delve' saw an accelerated adoption increase of 48%.

This trend suggests that humans are beginning to integrate characteristics of LLM-generated text into their own speech. According to a study contrasting linguistic patterns, LLM-generated text often features a more restricted vocabulary, more uniform sentence lengths, and generally exhibits less intense emotional expression, particularly fewer negative emotions like fear and disgust, compared to human-written text. LLMs also tend to utilize more numbers, symbols, and auxiliary verbs, contributing to a more objective and structured linguistic style.

The researchers from the arXiv study warn that these findings raise "societal and policy-relevant concerns about the potential of AI to unintentionally reduce linguistic diversity." They emphasize a "complex, bidirectional relationship in which humans and machines influence each other within a shared cultural environment," suggesting a future where human linguistic diversity could diminish as a result of this ongoing interaction with AI.