LLMs as Bags of Words

Treat chatbots not as minds but as giant 'bags' that return the most relevant word sequences from everything they’ve ingested. This explains weird outputs—hallucinated citations, automatic apologies, glue-on-pizza—without invoking intent or beliefs. It’s a practical mental model for predicting when they’ll be useful versus brittle. — A clearer public model of AI behavior curbs overtrust and anthropomorphic panic, guiding better product design, regulation, and everyday use.