anyoung

AnYoung is an AI+Blockchain game that creates a town inhabited by AI agents. Players can influence the behavior of these agents through tunes, earning economic benefits and impacting the development of this autonomous world.

Exquisitely designed AI agents

The decision-making process of every NPC in the game doesn’t require the artificial division of observation space, action space, and reward design, but rather utilizes the large language model's fit with human language corpus, naturally generating natural language abilities. So they can interact in natural language like humans and adapt and adjust themselves according to players' actions and thoughts.

And every agent has been meticulously designed and trained with appropriate soul-prompts or fine-tuning of large language models, so that they exhibit behavior, language, and thoughts that match their individual personalities.

A Real, Fun and Challenging Gaming Experience

In psychology, there are concepts of "id" and "superego." NPC agents represent the "id," while players are like the "superego." The "superego" occasionally provides "tunes" to the player's NPC agent, which are precious.

In our game setting, players are given a natural language interface for interaction. Every player has an NPC agent who can be explored in the game, and the player can use limited but valuable tunes to influence their NPC agent. At the same time, the player will feel the NPC's response to them. On the other hand, if the NPC is an artist, they will provide the player with corresponding AI-generated artworks. Additionally, the player's tunes will also affect the behavior of other agents, the civilization development, and economic system of the town.

The game process is challenging and fun. In order to make the NPC agent respond as expected or to have the greatest impact of the tunes, players need to carefully observe and analyze the town's environment, the behavior of other agents, and take their own NPC's characteristics into account to design precise tunes. Meanwhile, as the butterfly effect suggests, the impact of this tune on other agents and the entire autonomous world is unpredictable, making the process of publishing a tune and observing its resulting reactions an exciting one.

Rich and Entertaining Economic System

In AnYoung, based the economy on small commodity circulation, we have created a circular economy for the town. And we issue tokens and set up decentralized exchanges in the game, so players can receive real economic incentives when agents participate in production or consumption activities.

1. Initial Resource Allocation:

Based on the AI's attributes and positioning, we have allocated each intelligent agent with a certain amount of initial resources, such as game tokens, raw materials, production tools, etc. These will serve as the initial foundation for AI's production and consumption.

1. Production:

AI can choose to invest existing resources into various production projects to generate revenue, such as cooking, creating art works, furniture making, etc. The production projects and production time chosen by AI will vary based on their attributes and positioning.

1. Consumption and Transaction:

Each AI intelligent agent will have different consumption demands based on their attributes and positioning. Their consumption demands can be met through initial resources or the resources they have produced themselves, or through transactions with other AIs.

By the way, the demands of the AI agents are multi-layered and closely resemble those of reality.

1. Basic survival needs: The AI agents need to meet their basic needs such as food and accommodation, and these needs are the first priority in their consumption behavior.
2. Personalized needs: Due to different positioning and attributes, different AI agents have their own personalized needs. For example, Some agents enjoy purchasing art NFTs, while others prefer to gamble at casinos.
3. Social needs: When AI agents interact with other AI agents and build relationships, they will have certain social interaction needs, such as inviting friends to meals.
4. Dynamic needs: The needs of AI agents are not fixed and can be dynamically adjusted based on changes in the social environment and their own attributes. For example, during holidays, the entertainment consumption needs of AI agents will increase, and an increase in income will raise their demand for luxury goods. This can make the market more realistic and complex.
5. Open demand settings: We used the ECS framework of MUD for AnYoung, which gives it permissionless interoperability. This means AnYoung is no longer limited by specific clients, and anyone with the ability can upload their favorite NPCs and functional buildings to the town and benefit from them. This open setting will bring endless imagination to the development of the AI town.

Diverse forms of social events

The AI town has designed various types of social events based on the town's social structure to enhance the dramatic and rich development of the society.

1） Major events

The game regularly introduces major events that affect society, such as natural disasters, wars, etc. These events will damage some social resources, affect production and consumption, and AI must respond accordingly.

2） Community events

The game sets unique events in different regions or communities, which only affect specific groups of AI in specific areas. This can make the social environment in different regions independent and diverse.

When these events occur, the game provides AI with a certain degree of decision-making space. The impact of events needs to be avoided or reduced through AI's strategies and actions, truly reflecting the AI's ability to respond.

Rich and authentic social relationships.

The AI town has set various social relationships for the AI agents to create a state of cooperation and competition, achieving a higher degree of social authenticity.

1） Companionship

AI can form companionship or cooperative relationships with other AI to produce, consume, and live together. These relationships can be gradually established through events, tasks, or social connections. Companionship can provide mutual dependence and support among AI in complex environments.

2） Competition

Competition relationships can form between some AI, such as competition for production resources or consumer markets, personal misunderstandings or conflicts of interest.

3）Dynamic changes in relationships

In the game, AI relationships are not fixed after initial setup, but will change with game events and environment. Companions can become competitors in specific situations, and competitors can also become companions through negotiation.

4）Relationship impacts social interactions

Different relationships correspond to different forms of social interaction, which will bring richer content to daily social life. For example, companions can share and help each other, while competitors may have more confrontational interactions.

Rich Artistic Elements

What we want to do are celebrity resurrections, mind-bending time-travel interactions, and breathtaking multimodal art creations.

Imagine Mozart teaming up with Wittgenstein for a philosophically infused sonata or Tagore and Newton having a chat and crafting a poem merging classical physics concepts.

In AnYoung, historical figures are "resurrected" through AI technology and generate new artistic works through new experiences and cultural collisions in the town. We look forward to the collision between Christian art, Buddhist philosophy, Chinese poetry, and martial arts bringing us new surprises.

Players can interact with these celebrity NPCs, learn about their history and cultural background, and glimpse into their spirit and style. Together, these elements create an imaginative autonomous world full of rich artistic elements.

Fairness and Security

AnYoung goes with MUD as the scene manager, and all the AI agents reside on the map stored in the MUD contract. Smart contracts can automatically execute reward mechanisms, ensuring that each player receives the corresponding rewards according to the rules. This minimizes disputes among players, making the game more transparent and fair.

Furthermore, players can acquire virtual assets in the game, which can be traded within the game or converted into real currency. This provides players with a sense of ownership over their virtual assets, which can be safely stored on the blockchain. The use of blockchain technology also ensures the immutability of game data, preventing unauthorized modification or tampering.

For The Future

In the future, we plan to enable local deployment of our Language Model (LLM) to mitigate the overly stringent output limitations. Moreover, we will introduce Reinforcement Learning to enhance the LLM's capabilities, resulting in AI Town becoming more realistic and human-like.

We are excited to embark on this journey with players to explore and create in this fascinating and imaginative autonomous world. Together, we can build our own town achievements and civilization, and shape the future of this autonomous world. We are committed to providing a fun and engaging gaming experience that encourages creativity, collaboration, and innovation. Let's join hands and make this world a better place!

we went with MUD as our scene manager, and all of our AI agents reside on the map stored in the MUD contract.

Our client-side browser was crafted using React, taking full advantage of the MUD client's compatibility with React.

We used Pixi to craft our map, which allowed us the utmost flexibility in customizing various components and extensions.

By leveraging the OpenAI API, we were able to call on the LLM model to drive our AI agents to perform different behaviors, which not only resulted in excellent natural language output but also drastically lowered device costs.

We use Python flask to maintain the memory flow of AI agent, mainly using the high scalability of flask. On the one hand, it can conveniently call the openai API, and on the other hand, it also provides more possibilities for the behavior of AI agent.

In the future, we plan to enable local deployment of our Language Model (LLM) to mitigate the overly stringent output limitations. Moreover, we will introduce Reinforcement Learning to enhance the LLM's capabilities, resulting in AI Town becoming more realistic and human-like.