Stanford published a paper a few days ago detailing their work with Generative Agents and it sounds like tech that is destined to show up in video games.

…computational software agents that simulate believable human behavior. Generative agents wake up, cook breakfast, and head to work; artists paint, while authors write; they form opinions, notice each other, and initiate conversations; they remember and reflect on days past as they plan the next day.

If you’re a fan of The Sims, Stardew Valley, Dwarf Fortress, Rimworld, or Animal Crossing the description of what they created will probably excite you.

We instantiate generative agents to populate an interactive sandbox environment inspired by The Sims, where end users can interact with a small town of twenty five agents using natural language. In an evaluation, these generative agents produce believable individual and emergent social behaviors: for example, starting with only a single user-specified notion that one agent wants to throw a Valentine’s Day party, the agents autonomously spread invitations to the party over the next two days, make new acquaintances, ask each other out on dates to the party, and coordinate to show up for the party together at the right time.

You can see a demo of the simulation here and can play/pause it at will. Clicking on any of the agents lets you see what they’re doing. If you follow the “State details” link for any agent, you can see even more detail about them, including their personality, daily schedule, memory of previous actions, and conversations they’ve had with any other agent.

Open world settings are common in big budget games and it feels obvious that this tech will make it into those experiences. NPCs often follow simple, static routines in those titles despite developers describing their world as “alive.” Generative agents like this could deliver on that promise, guaranteeing unique, emergent experiences for every gamer.