General Intelligence

A framework for building self-organizing, reactive knowledge systems that learn, identify, and compose patterns through structural similarity.

Core Concept

Unlike traditional ML which requires fixed feature spaces, General Intelligence works with arbitrary nested data structures. Knowledge instances are active participants - they’re informed when learned, react to new knowledge, and can autonomously respond to events.

Key Features

• Hierarchical pattern matching: Find patterns within nested structures
• Active knowledge: Learned patterns react to new information and shape future learning
• Event-driven responses: Knowledge declares what it responds to and composes answers
• Autonomous operation: Knowledge can start monitoring and acting when learned
• Compositional reasoning: Combine knowledge from multiple sources
• Interpretable: Every comparison is based on explicit structural differences

Quick Example

from gi import GeneralIntelligence, Knowledge

# Create system and teach it patterns
gi = GeneralIntelligence()
gi.learn(Knowledge([1, 2, 3]))
gi.learn(Knowledge([Knowledge([1,2]), Knowledge([3,4])]))

# Identify similar patterns
matches = gi.identify(Knowledge([2, 3, 4]), threshold=5)
# Finds [1,2,3] with distance 3

# Hierarchical matching
matches = gi.identify(Knowledge([1, 2]), threshold=0)
# Finds [1,2] nested inside learned structure

Event-Driven Responses

class AlertKnowledge(Knowledge):
@classmethod
def is_response_for(cls, trigger, gi):
return trigger.get('type') == 'alert'

def compose(self, context, knowledge_class):
# Shape the response based on learned patterns
context['message'] = f"Alert: {context['trigger']}"

gi.learn(AlertKnowledge([...]))

# Trigger responses from matching knowledge
for response in gi.on({'type': 'alert', 'level': 'high'}):
print(response)

Autonomous Knowledge

class MonitoringKnowledge(Knowledge):
def is_active(self):
return True  # Start immediately when learned

def start(self, gi):
# Begin autonomous operation
self.monitor(gi)

def on_knowledge(self, new_knowledge, gi):
# React when new knowledge arrives
if self.should_alert(new_knowledge):
self.trigger_alert(gi)

Extending Knowledge

class PatternKnowledge(Knowledge):
def __init__(self, values):
super().__init__(values)
self.related_patterns = []

def on_knowledge(self, new_knowledge, gi):
# React when new knowledge is learned
if isinstance(new_knowledge, PatternKnowledge):
if self.similar_pattern(new_knowledge):
self.related_patterns.append(new_knowledge)
new_knowledge.related_patterns.append(self)

def compose(self, context, knowledge_class):
# Influence composition of new knowledge
context.update(self.contribute_to(context))

Philosophy

This models how knowledge actually works: not as passive data, but as active patterns that recognize, influence, organize, and respond. Knowledge knows the systems it belongs to, the other knowledge around it, and can act autonomously.

Use Cases

• Abstract reasoning (ARC challenge)
• Interpretable pattern recognition
• Event-driven AI systems
• Compositional learning
• Autonomous agents
• Structural transformation discovery

Status: Early prototype. Core functionality works. Community contributions welcome.