Non-playable characters (NPCs) are a thing of the past. Thanks to a groundbreaking effort by six computer scientists, NPCs can now join your quests, think for themselves, and take hints instead of just walking into walls. Their latest creation, “Smallville,” shows this leap in gaming technology (Park, et al., 2023).
Imagine a blend of Pokémon and The Sims, but with an IQ of 200. Smallville features 25 generative AI characters powered by a large language model (LLM) similar to ChatGPT. These characters, referred to as “agents,” can act, perceive, react, and interact with players and other agents to achieve goals. Here is a screenshot of the game:
Each character in Smallville has a detailed backstory and can draw conclusions about themselves, other agents, and their environment by storing new information in memory. They exhibit believable human behaviors—planning, engaging in meaningful conversations, and unwinding at a bar in the evenings (Works, 2024). For instance, when one character was tasked with hosting a Valentine’s Day party, they took the initiative to invite friends, decorate the café, and play matchmaker—all on their own! Below you can find an example of this conversation:
However, the scientists behind Smallville noted that these agents aren’t without quirks. Sometimes, they struggle to absorb information correctly, making choices like visiting closed stores or choosing a bar instead of a café for lunch (Park, et al., 2023). But isn’t that just a reflection of human nature? How often have you stood in front of a closed store or wanted a beer in the afternoon because it’s just one of those days?
This innovation excites me about the future of the gaming industry. Just think about the possibilities: Do these AI agents ponder the boundaries of their city? Do they wonder what lies beyond the tile maps? What if they tried to form their own government or even commit crimes? It would be fascinating to see how their interactions might mirror historical events, opening a landscape for storytelling as their population grows.
In my view, it would be engaging to create a game where these agents function under a set of unbreakable rules, allowing them to live their virtual lives independently of player involvement. Integrating each agent with an LLM could elevate the experience, enabling players to have dynamic, meaningful conversations that feel incredibly real. I’m eager to see where this technology leads us!
Check out the demo of the game here: https://reverie.herokuapp.com/arXiv_D…
Sources:
Works, F. (2024, 8 6). Why autonomous AI “agents” will be so important to your business. Retrieved from Fast Company: https://www.fastcompany.com/91167385/why-autonomous-ai-agents-will-be-so-important-to-your-business
Park, J. S., O’Brien, J. C., Cai, C. J., Morris, M. R., Liang, P., & Bernstein, M. S. (2023). Generative Agents: Interactive Simulacra of Human Behavior. Proceedings of the 36th annual acm symposium on user interface software and technology (pp. , 1-22.
I very much enjoyed reading this post as this topic is close to my heart. I cannot wait until technology like this is implemented into AAA gaming to allow for a truly immersive gaming experience. However, I wonder how far away we are from this technology being implemented into open-world games because of the current technical limitations. It is also interesting to ask questions on what limitations should be implemented on the ‘conscious’ agents to limit any offensive or bias in the agents but then as you mentioned they would be less human as biases and other ‘negative’ traits are part of human traits. Even though I believe this is the right direction for the future of gaming, I think a discussion about the extent of the consciousness of such agents is necessary.