Abstract
In this paper, we present God's Innovation Project (GIP), a god game where players collect words to dynamically terraform the landscape using generative AI. A god game is a genre where players take on the role of a deity, indirectly influencing Non-Player Characters (NPCs) to perform various tasks. These games typically grant players supernatural abilities, such as terrain manipulation or weather control. Traditional god games rely on predefined environments and mechanics, typically created by a human designer. In contrast, GIP allows players to shape the game world procedurally through text-based input. Using a lightweight generative AI model, we create a gamified pipeline which transforms the player's text prompts into playable game terrain in real time. To evaluate the impact of this AI-driven mechanic, we conduct a user study analyzing how players interacted with and experienced the system. Our findings provide insights into player engagement, the effectiveness of AI-generated terrain, and the role of generative AI as an interactive game mechanic.
Introduction
Procedural Content Generation (PCG) has long been a staple of game development, used to create vast, varied worlds. However, it is typically a "black box" process controlled by developers. While recent advances in Generative AI (GenAI) have introduced Procedural Content Generation via Machine Learning (PCGML), these tools are usually aimed at assisting designers rather than empowering players.
God’s Innovation Project (GIP) transforms generative AI from a background developer tool into a core, interactive gameplay mechanic. By integrating a lightweight text-to-image model into a "god game," we allow players to terraform their environment in real-time. This project explores whether GenAI can be reframed as a fun, functional game mechanic that enhances player agency and strategic depth, rather than just serving as a visual novelty.
The GIP Terraforming Pipeline
The core innovation of GIP is the Terraforming Pipeline, which translates player intent into 3D game geometry through a gamified loop. The system is built on several key technical and design pillars:
- The Model: We utilized the "Five Dollar Model," a lightweight, feed-forward convolutional neural network designed for low-dimensional image generation. It takes a sentence embedding as input and outputs a 10x10 grid of integers representing specific tiles.
- Gamified Prompting: To prevent "broken" or nonsensical prompts, players are restricted to a "word bank." Players must collect words (like "forest," "river," or "house") as resources by performing in-game tasks.
- Real-Time Execution: When a player selects a sub-grid and submits a "sentence" of collected words, the server returns a tile map that is instantly spawned into the 3D game world.
- Post-Processing: To ensure playability, the system performs "smoothing" on the AI's output—for example, connecting adjacent water tiles to form rivers—overcoming the 10x10 resolution limits of the base model.
Results and Player Insights
Our user study with 19 participants provided significant insights into how players interact with AI as a creative tool.
- AI as a Strategic Resource: Players did not just use the AI for aesthetics; they used it to solve problems. For example, because "house" tiles spawn new villagers, players strategically hunted for house-related words to grow their population and army.
- Constraints Breed Creativity: By restricting players to a word bank, the game avoided the "garbage in, garbage out" problem. This forced players to experiment with their limited vocabulary, leading 78.9% of players to report that they felt encouraged to "think outside the box."
- Semantic Alignment: Despite the model's simplicity, 84.2% of players felt the generated terrain was semantically consistent with their prompts. This suggests that for gameplay purposes, "vibe-accurate" generation is often more important than high-fidelity realism.
- User Sentiment: Integrating AI as a direct mechanic helped mitigate general negative sentiment toward the technology by giving the player direct, creative control over it.
Conclusion and Future Work
God’s Innovation Project demonstrates that generative AI can be a functional, engaging game mechanic that empowers players. While the current model has limitations in generalization, future work could involve using more robust models to allow for more diverse arrangements or multi-modal integration, such as AI-generated NPC dialogue that reacts to the player-created world. By giving players the "brush" to paint their own landscapes, we move toward a future where game worlds are as dynamic as the players' imaginations.
Summary updated