I find it more useful to think in terms of information theory and the related results in machine learning, signals, and computational theory. For example, I cla

I find it more useful to think in terms of information theory and the related results in machine learning, signals, and computational theory. For example, I claim that 'linguistic relativity', AKA the Sapir-Whorf hypothesis, is an informal intuitive restatement of some of the consequences of Kolmogorov complexity and other parts of algorithmic information theory and machine learning. Put simply: Different languages are differently efficient for expressing and therefore working with a given piece of information. The hypothesis space you explore is defined by the language you use to express hypotheses and a hypothesis space that is smoother and cleaner in the region of the "true" or "target" function will give more correct results more easily. In this case, confusing the meaning of the words that describe hypotheses near a particular target concept can force the use of a language that lacks terms that efficiently represent that concept, which dramatically hinders computations involving that concept. https://news.ycombinator.com/item?id=12110997 https://news.ycombinator.com/item?id=12110997