I’ve recently started to give a lot more credence to the weak sapir-whorf hypothesis—language not as a formal system that we use everywhere in the brain, but just as a high-level method we use to construct a distribution over lower-level sense data, which then affects perception via the top-down part of inference. The weirdness of color perception and language I think backs this up, but it’s also interesting to think about the reverse process of the language/model becoming richer like this. Humans are really amazing.
I’ve recently started to give a lot more credence to the weak sapir-whorf hypothesis—language not as a formal system that we use everywhere in the brain, but just as a high-level method we use to construct a distribution over lower-level sense data, which then affects perception via the top-down part of inference. The weirdness of color perception and language I think backs this up, but it’s also interesting to think about the reverse process of the language/model becoming richer like this. Humans are really amazing.