A.I. Agents
Each A.I. bot in ASM is called an Agent.
An Agent consists of:
A Brain NFT, which is the core of the Agent
A Form NFT, which encodes physical attributes (what it looks like, and how it operates)
Memories, stored in the Brain's Memory Tree, which encode behavioural strategies the Agent has learned (AI model training)
This model allows nearly unlimited flexibility in the creation of Agents for different worlds or use case.
It's not compulsory for all of these attributes to be used in every use case.
Some implementations may not need a Form, and attributes for the Brain may be unconstrained. But this structure will allow the introduction of randomness (and thus enhance the depth of) certain types of metaverse experiences. This is similar to how different genetics produce humans with different abilities, or different ways of solving similar problems.
1. Brain
The base NFT of every character. Each new Brain is created with a Genome Matrix - a randomized set of base values.
Genome Matrix values correspond to stats like strength, speed and size (in gaming) or risk tolerance and randomness (in finance).
The Genome Matrix uses the concept of a heat map grid for it's underlying values. This gives the Agent areas of strength and weakness. Different worlds can map to different areas of the agent's grid to "express" statistics. See Agent Parameters.
Brains also contain a Memory Tree - an immutable record of the memories created and owned by that Brain.
2. Memories
To improve in skill, Agents need to learn through experience (slow) or training (usually faster).
Memories store the outputs of Gym Training as machine learning models, in format like ONXX (or similar).
Memories are unique to Brains. Because they encode ML strategies, they are usually specific to the World the Agent operates in. However Memories from similar Worlds can be used as base training for another world that operates using the same or similar World Expression.
3. Form
The outer shell of an Agent... what it looks like and how it works. Forms are specific to a World (e.g. financial market, game arena).
The Form encodes unique visual elements, frequently 3D or 2D artwork, which can be upgraded, traded and replaced.
In physical activity-based environments like games and metaverses, Forms can enhance and modify Brain base values, and help define how it operates (e.g. wield a sword, kick a ball, pick up an object).
In non-physical environments, the Form is typically a unique, tradeable skin that helps identify or brand the Agent.
Forms are the way the Agent's Brain interacts with the world (similar to a sleeve in the TV show Altered Carbon). The way a Form modifies the performance of the Agent are defined by the World Owner, using ASM specifications. See Agent Parameters.
Examples of Brains, Forms and Memories creating an Agent
Game Character Agent
In the example of a game character, the Form (how it looks) and the Brain (its Genome Matrix and Memory Tree) are equally important. Both would guide the expression of the attributes, in conjunction with the expression model in the game world. See Agent Parameters.
DeFi Trading Agent
In the example of a DeFi Trading bot, the Form serves largely as a visual identifier (or a brand reference). Memories are hugely important, as they encode the strategies the Agent would use to place trades.
Other Use Cases
Altered State Machine Agents have a huge variety of potential use cases. These will continue to grow, as the infrastructure of Metaverses grows in connectivity.
Type | Examples |
Sports Games | Soccer, Rugby, or Tennis-playing Agents |
Combat Games | Robots, Spaceships, Superheroes |
Multiplayer Fantasy Games | Ice Zombie Armies, Elven Raiders, Orc Hordes |
Chatbots | Customer service agents, Smart personal companions (e.g. the OS in the movie Her) |
DeFi & Trading | Delta traders, Market Makers, Fund Management Agents |
Digital Assistants | A collective Siri that many people own, use and train |
Metaverses | Smart assistants, Personal representatives, or Digital friends in online worlds (e.g. Ready Player One or Snow Crash) |
Last updated
