In the past year, AI agents based on large language models (LLMs) have rapidly become one of the most exciting yet controversial topic. Some think it's the next big thing, while others think these agents are just thin wrappers around LLMs. In this tutorial, we hope to carefully examine and reconcile the different viewpoints and properly contextualize the new generation of AI agents in the broader history of AI. We believe contemporary AI agents are qualitatively different from the previous generations (e.g., logical agents or neural agents). By integrating LLM(s), they gain a new capability of using language as a vehicle for reasoning and communication, which substantially improves their expressiveness and adaptivity. Therefore, they are best called language agents, for language being their most distinct trait.
Language played a crucial role in human cognitive evolution, and AI might be following a similar path. However, there has been little systematic discussion on the definition, theoretical foundations, applications, risks, and future directions of language agents. This cutting-edge tutorial aims to fill that gap by providing a comprehensive exploration of language agents.
Note that this tutorial is not meant to be a comprehensive survey of related work or a practitioner's guide that focuses on code frameworks.
Our tutorial will be held on November 15 (all the times are based on EST = Miami local time).
| Time | Section | Presenter |
|---|---|---|
| 14:00—14:20 | Part I: Introduction [slides] | Yu Su |
| 14:20—15:20 | Part II: Foundations: Reasoning, Memory, and Planning [reasoning+memory] [planning] | Shunyu Yao, Yu Su |
| 15:20—15:30 | Q&A Session I | |
| 15:30—16:00 | Coffee Break | |
| 16:00—16:45 | Part III: Applications, Data, and Evaluation [slides] | Tao Yu |
| 16:45—17:20 | Part IV: Emerging Topics: Multi-Agent Systems, Safety, and Social Impact [slides] | Diyi Yang |
| 17:20—17:30 | Part V: Final Remarks and Outlook + Q&A Session II [slides] | Diyi Yang |
@inproceedings{language-agent-tutorial,
title = "Language Agents: Foundations, Prospects, and Risks",
author = "Su, Yu and Yang, Diyi and Yao, Shunyu and Yu, Tao",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Tutorial Abstracts",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-tutorials.3",
pages = "17--24",
}