First, there is no universally agreed-upon test to measure sentience. So I don't see much value in debating whether LLMs are sentient or not... but they aren't

First, there is no universally agreed-upon test to measure sentience. So I don't see much value in debating whether LLMs are sentient or not... but they aren't ;) Second, the similarity of deterministic algorithms' outputs to what humans consider expressions of sentient beings is not an argument in favor of LLMs being sentient. It's just like arguing that a magic trick must be real because it looks so much like the real thing and not a trick.