The article argues that large language models (LLMs) are fundamentally cultural artifacts demanding new societal scripts, not the human-like entities or conscious threats often imagined. Cultural scripts for interacting with these unpredictable, non-sentient text generators are urgently needed. The author critiques prevailing myths—like Star Wars droids or conscious AIs—as inadequate for understanding LLMs' unique, alien behavior. Instead, concepts like Searle's Chinese Room or Chalmers' philosophical zombie offer better analogies for their opaque internal processes.

This cultural gap risks poor decisions, from mandating LLM use to trusting their summaries of sensitive data. The piece warns that big tech's control over these models could significantly influence public expression, much like social media algorithms have marginalized certain voices.