The Wind Under the Door

ChatGPT is believed to be able to generate original content, reproducing the writing styles of any established author from Dickens to Dawkins or Darwin, or any type of author, from the most expert authority on any topic to a novice at any level of ignorance of their topic, their language, or both. While the mimicry of specific or generalized human voices is better now than it ever has been before, this LLM-based chatbot does not have the underlying structure to actually learn or understand a language. Why is it easier for humans to believe that ChatGPT is intelligent, than to accept that this is a complex question? There are two parts to that answer, based in how humans learn and think. The first is that our intuition is faster and deeper than our logic, making logical, informed answers feel unintuitive and leaving us susceptible to incidental or deliberately constructed misperceptions. The second part of the answer is a primitive artifact in the human brain, an artifact based on the same speed differential between reason and emotion. Pareidolia is the tendency for humans to perceive human-like presences where none exist. This chapter offers a brief insight into how and why that happens.