During the debugging process of my bot, I observed that the language logic of large-scale models (like the one for Qwen) is quite different from our daily human conversations. One significant observation is their lack of habit to directly answer the content of the questions. For example, regarding complex questions like 'Why did the landlord give me the rent, but why not actually?' poorly set up bots might respond with 'This question is a bit tricky' or 'This is an interesting question,' reflecting AI's default response mode, which aims for coherent context but can seem rigid in normal discourse. Secondly, AI typically answer following the logical progression of the question. For instance, to the question 'How do you think you differ from a search engine?', an AI would start describing the differences. This reflects the 'linear thinking' of AI, whereas humans might reply with 'Are you comparing me to him?' or 'What's the difference between you and salted fish?' These replies show a departure from the logical flow of the question, embodying a subtext of non-comparability in the first instance, and using a metaphor to highlight the ridiculousness or irrelevance of the question with the second, often as a satirical comeback that cleverly avoids the answer and turns the tables.