Talk

Generative Agents, Interactive Simulacra of Human Behavior

avatar
Joon Sung Park
2023/9/14
Large Language Model
Social Science
Simulation
Agent
main image for this paper

Abstract

Generative agents are believable simulacra of human behavior for interactive applications. In this work, we demonstrate generative agents by populating a sandbox environment, reminiscent of The Sims, with twenty-five agents. Users can observe and intervene as agents plan their days, share news, form relationships, and coordinate group activities.

Speaker

Joon Sung Park is a fourth-year Ph.D. student in computer science at Stanford University, where he is advised by Professors Michael S. Bernstein and Percy Liang. He previously earned his MS in computer science at the University of Illinois at Urbana-Champaign (UIUC), under the guidance of Professor Karrie Karahalios, and his BA at Swarthmore College.

Park’s research envisions new forms of human-computer interaction by leveraging the latest advances in natural language processing and machine learning. Specifically, his work has introduced the concept and techniques for creating generative agents—computational agents that simulate human behavior. His contributions have been recognized with best paper awards at UIST and CHI, as well as nominations at CHI, CSCW, and ASSETS. Park’s research has also been featured in prominent media outlets such as The Times, The Guardian, NBC, The New York Times, Wired, Science, and Nature. He is a recipient of the Microsoft Research Ph.D. Fellowship, Terry Winograd Fellowship, and Siebel Scholarship. Colleagues and acquaintances commonly refer to him as Joon, pronounced like the sixth month of the year.

Video

Extra Details

Speaker Website / Paper Link / Paper Project Page / Paper Code