Large-scale generation of meaningless conversation data may be required in certain specific research or testing scenarios, but attention should be paid to its use and potential impact. Some natural language processing tools and frameworks can assist in generating simulated conversation data. For example, the text generation ability of language models can be used to construct some seemingly illogically related sentences to form conversations. Open source language models such as GPT-2 based on the Transformer architecture can generate a large amount of text by adjusting input prompts and parameter settings, from which they can be screened or further processed into simulated conversations. However, there may also be some problems with the random generation of large amounts of meaningless data, such as uneven data quality, lack of practical value and semantic coherence. At the same time, the generated data should be used within a legal, compliant, and ethical framework.