I love this theory. You're saying that a distinction can be drawn between our linguistic concepts and our lived experience, and that the former can be learned w

I love this theory. You're saying that a distinction can be drawn between our linguistic concepts and our lived experience, and that the former can be learned without the latter. And that a model could operate upon those linguistic concepts in a useful way, but without the benefit (or drawback?) of the mappings we keep between language and experience. And that it can learn this based on the large amount of texts we have. Fascinating, and seems like a plausible description of what's going on.