Scholarlink

<p>Ian McEwan’s “Machines like me” presents a scenario where androids demonstrate a limited understanding of the world. In this study, we feed the content of “Machines like me” novel into a chatbot powered by LLMs to analyze how it interprets various questions posed as prompts. We create a private chatbot based on OpenAI API using in-context learning and LangChain orchestration for evaluation. Our focus is on assessing the chatbot’s accuracy in responding, its level of world understanding, and the consistency of its answers. Our findings reveal that chatbots, similar to the android in the novel, still exhibit a lack of world understanding. This limitation, coupled with their tendency to produce hallucinatory responses, can hinder their ability to correctly interpret and respond to textual input. Further, we compare the private chatbot with ChatGPT-4 and evaluate the results.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Interpreting text corpora from androids-related stories using large language models: “Machines like me” by Ian McEwan in generative AI

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年3月6日

查看全文
  • Simona-Vasilica Oprea,
  • Adela Bâra

摘要

Ian McEwan’s “Machines like me” presents a scenario where androids demonstrate a limited understanding of the world. In this study, we feed the content of “Machines like me” novel into a chatbot powered by LLMs to analyze how it interprets various questions posed as prompts. We create a private chatbot based on OpenAI API using in-context learning and LangChain orchestration for evaluation. Our focus is on assessing the chatbot’s accuracy in responding, its level of world understanding, and the consistency of its answers. Our findings reveal that chatbots, similar to the android in the novel, still exhibit a lack of world understanding. This limitation, coupled with their tendency to produce hallucinatory responses, can hinder their ability to correctly interpret and respond to textual input. Further, we compare the private chatbot with ChatGPT-4 and evaluate the results.