Abstract
Despite the astonishing level of intelligence demonstrated by large models such as ChatGPT, they still have inherent limitations beyond the realm of text. Therefore, the emergence of intelligent agents based on large models and capable of using external tools has emerged. Big model intelligent agents are committed to building task oriented digital experts or teams. These agents use language as a knowledge carrier and can independently explore or collaborate to produce complete solutions when human users set task requirements. This research direction has opened up new opportunities for the automation of complex problem solving and has significant potential to improve the production efficiency of “intelligent agent collaborative human work”, thereby freeing human labor from heavy labor. This report will delve into the core dimensions of large model intelligent agents, with a focus on cutting-edge developments in construction, collaboration, evolution, and expansion, as well as case studies in education, military, embodied, and end-to-end scenarios. The aim is to promote the development and application of autonomous and efficient large model systems.
Speaker
Qian Chen, a postdoctoral fellow at the Natural Language Processing Laboratory of Tsinghua University and a scholar of water and wood at Tsinghua University. My main research directions are natural language processing and large-scale model swarm intelligence. I have published several papers as the first author in international academic conferences or journals related to artificial intelligence, information management, software engineering, and other fields. Led and released ChatDev, a collaborative software development framework for large model intelligent agents, in 2023, and has topped Github Trending charts multiple times.