Essay.05-ZhuoyingLi-2100017764
Keywords: language model, communication, shared experience, cooperativeness
Abstract: The recent proliferation of Large Language Models (LLMs) such as GPT-4 underscores their enormous potential for a wide range of applications, including drafting legal documents, aiding communication, and even creative writing. Despite the impressive capabilities of these models in generating human-like text, there is a recognized deficiency in comprehensive testing frameworks that assess their true communication abilities. This research delves into the core aspects of human communication, focusing on the unique markers such as the common conceptual ground and prosocial motivation. It raises critical questions about whether these models truly understand and communicate information or are simply mimicking linguistic patterns. The essay then presents a set of proposed tasks aimed at evaluating LLMs' ability to utilize common conceptual ground and engage in multi-agent coordinative communication.
Submission Number: 116
Loading