I wonder how the nature of the language used to train an LLM affects its model of the world. Would a language designed for the maximum possible information content and clarity like Ithkuil make an LLMs world model more accurate?