Agents Thinking Fast and Slow: A Talker-Reasoner Architecture

This paper proposes a Talker-Reasoner architecture inspired by the "Thinking, Fast and Slow" framework. It separates an AI agent into two components: a fast, intuitive "Talker" module for rapid responses, and a slow, deliberate "Reasoner" module for deep reasoning. This dual-system approach aims to balance efficiency and accuracy in AI agent interactions.