Unbounded: A Generative Infinite Game of Character Life Simulation
Jialu Li, Yuanzhen Li, Neal Wadhwa, Yael Pritch, David E. Jacobs, Michael Rubinstein, Mohit Bansal, Nataniel Ruiz
2024-10-25
Summary
This paper introduces Unbounded, a new type of video game that uses advanced AI to create an endless character life simulation where players can interact with their characters in a highly dynamic and personalized way.
What's the problem?
Traditional video games are often limited by fixed rules and pre-coded scenarios, which restrict player creativity and the variety of experiences. Players want more freedom to shape their gameplay and have unique interactions with characters in a game world that feels alive and responsive.
What's the solution?
Unbounded uses generative models to create a game where everything is generated on-the-fly based on player interactions. It features a specialized large language model (LLM) that generates game mechanics, stories, and character interactions in real time. Players can customize their characters and guide them through a virtual world, with the game adapting to their choices. The system also includes a new visual component that ensures characters look consistent across different environments.
Why it matters?
This research is significant because it represents a major step forward in gaming technology, allowing for more immersive and interactive experiences. By using generative AI, Unbounded opens up new possibilities for how games can be designed, making them more engaging and personalized for players. This approach could change the future of gaming by creating worlds that evolve based on individual player experiences.
Abstract
We introduce the concept of a generative infinite game, a video game that transcends the traditional boundaries of finite, hard-coded systems by using generative models. Inspired by James P. Carse's distinction between finite and infinite games, we leverage recent advances in generative AI to create Unbounded: a game of character life simulation that is fully encapsulated in generative models. Specifically, Unbounded draws inspiration from sandbox life simulations and allows you to interact with your autonomous virtual character in a virtual world by feeding, playing with and guiding it - with open-ended mechanics generated by an LLM, some of which can be emergent. In order to develop Unbounded, we propose technical innovations in both the LLM and visual generation domains. Specifically, we present: (1) a specialized, distilled large language model (LLM) that dynamically generates game mechanics, narratives, and character interactions in real-time, and (2) a new dynamic regional image prompt Adapter (IP-Adapter) for vision models that ensures consistent yet flexible visual generation of a character across multiple environments. We evaluate our system through both qualitative and quantitative analysis, showing significant improvements in character life simulation, user instruction following, narrative coherence, and visual consistency for both characters and the environments compared to traditional related approaches.