This is an orthogonal question to whether LLMs have qualia (almost certainly no) and to the question of whether any hypothetical qualia would be in any way correlated with word choice (also almost certainly no), as opposed to mechanistic factors such as runtime and memory access patterns.
Source: [Hacker News](https://news.ycombinator.com/item?id=48945125)