I tend to think of it as reappropriating words in a different context. Since we're talking about language models, they're analogues but not as we would assign the same meaning to other humans.