This is a very interesting use case of the Generative Agents project.
I feel like this is just one of the earlier attempts at creating an interactive experience with this, but there's still so much potential for future games, if the technology is refined further. Think of something like Animal Crossing (or any other game where interacting with NPCs is at the forefront of gameplay), but with the characters that have completely unique personalities and an ability to generate new dialogue (rather than relying on a limited number of pre-written lines), or react to player actions in unique ways, and so on.
I feel like this might be the beginning of a much richer single player game genre that mixes generative language, image, texture, and other features with procedural rogue like world generation with the semantic features defined by the LLM. Obviously NPCs, dialogues, etc, can all be context aware agents bootstrapped by the environment and player history. Content creators IMO will be creating assets that are used to fine tune and create LoRas as future game assets. Instead of monotonously creating every asset in a game, they would create styles and control nets, language corpuses, and human written lore RAG’ing into game context, etc. I think artists will have a wonderful contribution much richer than simply painting a picture.
I feel like this is just one of the earlier attempts at creating an interactive experience with this, but there's still so much potential for future games, if the technology is refined further. Think of something like Animal Crossing (or any other game where interacting with NPCs is at the forefront of gameplay), but with the characters that have completely unique personalities and an ability to generate new dialogue (rather than relying on a limited number of pre-written lines), or react to player actions in unique ways, and so on.