The Secret Way ChatGPT Is Changing How We Talk

About this video

• Video Title: The Secret Way ChatGPT Is Changing How We Talk
• Channel: Taylor Lorenz
• Speakers: Taylor Lorenz, Adam Elkus (Guest)
• Duration: 00:26:51

Overview

This video explores how ChatGPT and other large language models (LLMs) are subtly influencing human language, both in written and spoken forms. It discusses specific word usage changes, the concept of "AI-speak," and the potential for linguistic homogenization, while also considering how humans might react and adapt to these changes.

Key takeaways

• ChatGPT's Linguistic Impact: ChatGPT, since its 2022 launch, has rapidly become embedded in daily life, influencing not only how we discover information but also how we write and speak.
• Shifting Vocabulary: Research shows a significant increase in the use of words like "delve" and "tapestry" in written communication following ChatGPT's release, indicating a direct impact on vocabulary.
• "AI-Speak" and Avoidance: A distinct way of communicating, termed "AI-speak" or "LLM-speak," is emerging, characterized by specific word choices and sentence structures. This has led to "avoidance speech," where people consciously try not to sound like AI.
• Technology's Historical Influence on Language: Language has always evolved in response to new technologies, from the telephone popularizing "hello" to autocorrect affecting spelling. AI is the latest technological influence.
• Concerns of Homogenization and Authenticity: There are concerns that AI could lead to a homogenization of language, reducing creativity and distinctiveness. Conversely, there's a growing desire to express authenticity and uniqueness in communication, leading to trends like "Italian brain rot" as a human reclamation of AI.
• AI as a Tool vs. Cognition: While AI can be a useful tool for editing and finding new ways to express oneself, relying on it for cognition raises concerns about losing autonomy in thought and speech.
• Bias in AI Training Data: The way AI models are trained, for example, the disproportionate use of "delve" by West African trainers, can lead to biased language patterns that are then reflected back to users, potentially leading to unfair accusations of "sounding like AI."
• Personification of AI: People tend to personify AI, treating it like a human, which influences how they communicate with it and how the AI, in turn, responds. This can obscure the complex technological process behind AI.