The LLM as a modern Ouija board. Many hands on the planchette. The collective weight of training data — billions of human utterances — moves the output. No single participant steers it. The result feels like contact with something, but it is the pressure of the crowd refracted through statistics.
What does it mean to ask a question and receive an answer that is the averaged intention of millions of anonymous writers? Is it wisdom, or noise shaped into confidence?