2 entries
Three soft layers between a user and hallucination all fail silently. Three hard defenses make confabulation visible, measurable, and refusable without suppressing the answer.
Argues that the appeal of large language models is rooted in human interactional and interpretive processes, not in machine intelligence. Tracing a line from divination and ELIZA to present-day LLMs, Dingemanse follows Lucy Suchman in slowing down 'discourses of the smart machines' and shows how fluid output, fine-tuned overconfidence, and interactive design exploit our interpretive infrastructure. Critical AI literacy, the paper argues, must be grounded in a deep understanding of human interaction and sense-making.