Experiments

Toy Reasoning System

A step-by-step logic demo in MeTTa

Toy Reasoning System

This experiment demonstrates how to build a basic inference engine using MeTTa's pattern-matching capabilities.

The Knowledge Base

We start by populating our AtomSpace with some simple categorical relationships.

(IsA Eagle Bird)
(IsA Sparrow Bird)
(IsA Penguin Bird)
(Properties Bird (CanFly True))
(Properties Penguin (CanFly False))

The Inference Engine

We want to find out if an animal can fly. The engine needs to check for specific properties first, and then fall back to general category properties.

(= (can-fly $animal)
   (let $specific (match (AtomSpace) 
                        (and (IsA $animal $type) (Properties $animal (CanFly $val))) $val)
        (if (empty $specific)
            (match (AtomSpace) 
                   (and (IsA $animal $type) (Properties $type (CanFly $val))) $val)
            $specific)))

Running the Demo

Test Eagle: !(can-fly Eagle) -> Result: True (Inherited from Bird).
Test Penguin: !(can-fly Penguin) -> Result: False (Overridden by specific property).

Flow Diagram

graph TD
    Query[Can animal X fly?] --> Specific{Specific Prop?}
    Specific -->|Yes| Out1[Return Specific Val]
    Specific -->|No| General[Check Category Prop]
    General --> Out2[Return General Val]

Key Insight

This "Toy" system demonstrates Default Reasoning and Inheritance, which are core components of sophisticated AGI systems like Hyperon.

Next: Reinforcement Learning Demo

Building Simple Agents

Practical blueprints for reactive and deliberative agents

Reinforcement Learning Demo

Simple RL loop within the AGI framework

On this page

Toy Reasoning System The Knowledge Base The Inference Engine Running the Demo Flow Diagram Key Insight