We describe our reality using words. Large Language Models (LLMs) create an abstraction of our use of words, not of reality. The output of these models can be garbled (an hallucination) as a result of this difference.