> Anthropomorphising them or their software programming is harmful¹. LLMs are trained on internet data produced by humans. Humans tend to appreciate politenes

> Anthropomorphising them or their software programming is harmful¹. LLMs are trained on internet data produced by humans. Humans tend to appreciate politeness and go to greater lengths answering polite questions, hence the LLMs will also mimic that behavior because that's what they're trained on.