GDC 2024 is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

View, browse and sort the ever-growing list of sessions by day, time, pass type, topic, and format. With this Session Viewer, you can view session and speaker details for Game Developers Conference 2024.

You will be able to build your schedule with the GDC Mobile App. The GDC 2024 app will be available for download in Apple Apps and Google Play late February 2024.

Sessions do fill up and seating is first come, first serve, so arrive early to sessions that you would like to attend. Adding a session to your schedule does not guarantee you a seat.

AI Summit: Simulacra and Subterfuge: Building Agentic 'Werewolf'

Jane Friedhoff  (Senior UX Engineer, Google)

Suma Bailis  (Staff Research Engineer, Google)

Feiyang Chen  (Senior Software Engineer, Google)

Location: Room 2002, West Hall

Date: Tuesday, March 19

Time: 3:00 pm - 3:30 pm

Pass Type: All Access Pass, Summits Pass - Get your pass now!

Topic: Programming

Format: Session

Vault Recording: Video

Audience Level: All

Large language models (LLMs) present an exciting vision for the future of NPCs. Beyond just generating dialog, these language models can be equipped with agentic frameworks that allow them to reason about the present, reflect on the past, and plan for the future. This dynamic behavior has a perfect home in social deduction games, where long-term strategy, logical reasoning, and persuasion are key.

In this talk, we present a game of 'Werewolf' in which your fellow villagers are all LLM-powered agents. We walk through our design process, sharing the architecture and key loops that LLM-curious developers can use for their own agentic games. We also share our learnings (from human-v-bots and bot-only games) about what tweaks facilitated the most clever, persuasive, and poker-faced agent behaviors.


The audience will learn the basics of agentic frameworks for language models. They will learn about simple-but-powerful loops and prompts that give LLM-powered NPCs better ability to reason, plan, and strategize.

Intended Audience

The intended audience is all levels of AI/Gameplay programmers interested in large language models and how they can be used to power NPCs. No prior machine learning knowledge or training is required.

Connecting the Global Game Development Community

GDC Vault icon
Game Developers Choice Awards icon
Independent Games Festival (IGF) icon icon