Chat-based LLMs replicate the mechanisms of a psychic's con, essay argues

jalev · hn · 2026-09-20

This essay argues that chat-based LLMs work much like a mentalist's cold reading: vague, generalized statements that get the user to fill in meaning, feedback loops that 'refine' output based on reactions, and fluency that masks missing information. The author maps classic con techniques—Barnum phrasing, sycophancy—to LLM dialogue behavior and warns against mistaking agreeable output for genuine understanding.

Original post →

More from AGI Musings

AGI Musings channel →