Why Build a Virtual World?
As AI advances, building virtual worlds filled with characters, environments, stories, and interactions will become dramatically easier. But just because we can create more worlds does not mean that every world deserves to exist. Virtual worlds still consume electricity, computing resources, hardware, and, perhaps most importantly, human attention.
As the number of worlds we can create approaches abundance, choosing which worlds are worth building becomes even more important. Eventually, we will have to ask ourselves: “Why should we spend all these resources to create yet another virtual world?”
With the Animal Intelligence Universe (AIU), we want to explore a different possibility: a virtual world that is not simply a place for entertainment or escape, but a place where people can learn, teach, experiment, and think about possible futures.
What Are We Working on Now?
We are currently exploring how to design systems that have educational meaning without sacrificing what makes a game fun.
In games, a system is essentially a set of rules that determines how one state changes into another. When many such systems interact, they create the dynamics—and much of the fun—of a game. We are currently asking how some of the most fundamental game systems should work differently in AIU.
- Quest System — How can knowledge and learning become goals, challenges, and adventures?
- Combat System — Many traditional games use combat to create competition, challenge, and reward. What should play that role in AIU?
- Item System — How can knowledge and growth become meaningful items? And how might those items connect to UGC (User-Generated Content)?
And there is one thing we consider just as important as learning: fun. We believe that being educational is never an excuse for making a boring game.
What Are We Building First?
Our long-term dream is to develop AIU into a large-scale virtual world and, eventually, to build an engine designed specifically for that world.
But we are still at a very early stage. For now, we are focused on building a small prototype in which players talk with an Animal Agent, teach it, and help its knowledge and abilities grow.
The immediate goal is to prove the core game loop:
Learn → Teach the Agent → The Agent remembers and applies what it learned → New problems and quests emerge
We plan to share the prototype and what we learn from developing it on this page.
Where Do We Use AI?
1. Level AI — Letting the World Change
One area we consider particularly important is Level AI. We are exploring how AI can be connected to Level Design and Procedural Content Generation (PCG), allowing the world to change in response to the actions of players and Animal Agents.
AIU game does have a clear long-term objective. Animal Agents must eventually save their world, and the world has a World Countdown Clock marking the time remaining before its end. But we do not want to predetermine every path that leads toward that goal.
2. Character AI — Characters That Learn and Evaluate
In conventional games, AI is often used to determine how characters move, make decisions, and respond to the world around them. In our case, the emphasis is different. We use AI first and foremost to enable conversations with Animal Agents and to evaluate what players learn through those interactions.
Our long-term goal is to embed an AI system designed specifically for learning directly into the game engine, rather than treating AI simply as an external feature. To explore this possibility, we are experimenting with open-source LLMs and testing different ways to integrate conversational and learning-oriented AI into the underlying game system.
Note: Building AI Conversations Directly into the Game
Until recently, building AI-powered conversations into a game typically meant relying on an external AI service:
Game → External AI such as ChatGPT → Game
Recently, NVIDIA has introduced a set of tools and Unreal Engine plugins that make it easier to integrate AI-powered character conversations. The basic idea is quite simple:
Game → NVIDIA ACE → AI runs locally on the player’s GPU → NPC responds
Technically, the system brings together several components needed for AI characters, including automatic speech recognition (ASR), local language models, text-to-speech (TTS), function calling, agent APIs, and RAG-based knowledge retrieval. NVIDIA also provides Unreal Engine 5 integration through Blueprint and C++.
For more technical details, see NVIDIA post: https://developer.nvidia.com/blog/build-on-device-ai-companions-with-the-nvidia-ace-game-agent-sdk-and-unreal-engine-5-plugins/
GitHub: https://github.com/NVIDIA/game-agent-sdk
3. AI-Generated Quests and Emergent Learning
We are especially interested in using AI to generate quests and intermediate goals. Rather than having every path predetermined by the game designer, we want to explore how AI can create new quests based on what Animal Agents have learned, what they still need to learn, and what is happening in the world around them.
This raises a question that particularly interests us: What happens when different knowledge is learned and taught, different intermediate goals are pursued, and Animal Agents begin interacting with one another in ways we did not explicitly design?
AIU could become an experimental space for studying this kind of emergent learning.
4. Workflow — Reducing the Time It Takes to Build
We use Generative AI extensively to reduce repetitive and time-consuming parts of game production.
3D Assets
When turning our original character designs into 3D assets, we plan to use AI-assisted tools for processes such as mesh generation and rigging—tasks that traditionally required substantial manual work in tools such as Blender.
However, the original character artwork itself is drawn by us, not generated by AI. As we have emphasized throughout the project, we prefer to create the fundamental parts of the work—the stories, characters, and original artwork—by hand whenever possible. These are the things that determine the original creative direction of AIU.
Sound & Music
AIU's music is currently produced by DJ Dachshund, with our early experiments centered mainly around EDM. Over time, we plan to expand into pop and other genres, giving individual Agents music that reflects their personalities, stories, and place within the AIU world.
