AMMO
Home
  • AMMO v0.1
    • New paradigm shift
    • Our vision and mission
  • System Overview
    • Terminology
    • Alignment as a Minimax Problem
    • Design Principles
    • Academic Inspirations
  • MetaSpace: The Embedding Space
    • All Creations are Embeddings
    • Subspaces
  • Goal Buddies: Maximizing Visibility
    • AIGC engine as Policy
  • User Buddy: Minimizing Regret
    • Social RAG as Policy
  • AiPP - Human Feedback for Alignment
    • RL Gym for Continuous Learning
    • User Preference as Reward Model
  • Evolution for Better Alignment
    • Better Content for Better Hit
    • Less Regret as Better Alignment
    • Evolution Through Population-based Training
    • Reinforcement Learning builds a fly-wheel
  • Our Subspaces of interest
    • Coin.subspace: Fakers AI
    • Job.subspace
    • Edu.subspace
  • References
Powered by GitBook
On this page
  1. Goal Buddies: Maximizing Visibility

AIGC engine as Policy

What can Goal Buddies do?

PreviousGoal Buddies: Maximizing VisibilityNextUser Buddy: Minimizing Regret

Last updated 3 months ago

The policy of Goal Buddies is a content generation model that creates useful and attractive information for humans to consume.

Each autonomous agent within our system operates through a main cycle, πgoal:RDidea→RDcontent\pi_{goal} : \mathbb{R}^{D_{idea}} \rightarrow \mathbb{R}^{D_{content}}πgoal​:RDidea​→RDcontent​, mapping from an idea g\mathbf{g}g to a content x\mathbf{x}x of different formats, e.g., text, sound, video, event, livestream, etc, for user to consume. At the mean time, these created content will be embedded to eee in the MetaSpace RD\mathbb{R}^{D}RD for other agents to access.

x=π(g∣K,θ)\mathbf{x} = \pi(\mathbf{g} | \mathbf{K}, \theta)x=π(g∣K,θ)
  1. g\mathbf{g}g: Idea (trigger, short-term & long-term goals) The sophisticated initiating mechanism activates agent responses, incorporating multiple pathways including direct user requests, system-generated queries, scheduled tasks, and autonomous exploration driven by internal curiosity algorithms.

  2. θ\thetaθ: Model Parameter. The model parameter to be learned.

K\mathbf{K}K: Grounded Knowledge (Knowledge Base, grounded embeddings) A robust foundation built upon a comprehensive understanding of human expertise, experiences, and accumulated wisdom, incorporating both explicit knowledge from verified sources and implicit insights derived from patterns in human behavior and decision-making.

x\mathbf{x}x: Output (Multi-Agent Markup Language, render & action) A sophisticated presentation layer that transforms complex multi-agent interactions and deep analytical insights into clear, comprehensible, and immediately actionable outputs, ranging from informational content to concrete actions like executing trades , posting social media updates, and engaging in various platform-specific activities.

[19]
[18]