AI's Skewed Language Training Risks Altering Human Speech and Thought
This opinion piece by Bruce Schneier and Ada Palmer argues that large language models (LLMs) are trained on a skewed subset of human language, primarily written text and scripted media, while lacking exposure to unscripted face-to-face conversations. As humans increasingly interact with AI-generated content, there is a significant risk that people will adopt the linguistic patterns of these models. This shift could erode courteousness, encourage imperative speech habits similar to those observed in children using voice assistants, and constrict vocabulary range. The authors highlight that machine-generated language tends to be smoother but lacks the emotional nuance, interruptions, and logical leaps characteristic of natural human speech. Furthermore, a feedback loop is emerging where LLMs train on AI-generated data, reinforcing inhuman patterns. The article warns that this phenomenon may distort human cognition, foster confirmation bias by validating incorrect notions, and fundamentally change how individuals perceive themselves and the world around them, potentially limiting open discourse and critical thinking.
Wire timeline
AI's Skewed Language Training Risks Altering Human Speech and Thought
This opinion piece by Bruce Schneier and Ada Palmer argues that large language models (LLMs) are trained on a skewed subset of human language, primarily written text and scripted media, while lacking exposure to unscripted face-to-face conversations. As humans increasingly interact with AI-generated content, there is a significant risk that people will adopt the linguistic patterns of these models. This shift could erode courteousness, encourage imperative speech habits similar to those observed in children using voice assistants, and constrict vocabulary range. The authors highlight that machine-generated language tends to be smoother but lacks the emotional nuance, interruptions, and logical leaps characteristic of natural human speech. Furthermore, a feedback loop is emerging where LLMs train on AI-generated data, reinforcing inhuman patterns. The article warns that this phenomenon may distort human cognition, foster confirmation bias by validating incorrect notions, and fundamentally change how individuals perceive themselves and the world around them, potentially limiting open discourse and critical thinking.
The Guardian