Description
A Minecraft 1.20.1 Forge mod that uses actual machine learning to make ALL vanilla mobs progressively smarter over time. Mobs learn from every interaction using a Deep Q-Network and evolve their behavior to create an ever-changing, adaptive world.
Features:
🧠 Real Machine Learning Deep Q-Network with experience replay Progressive evolution - mobs genuinely learn and improve Online training - neural network trains during gameplay Model persistence - learned behavior saves across sessions World-adaptive - discovers tactics that work in YOUR world 70+ mob types - Every vanilla Minecraft mob can learn and evolve
🎮 Universal Mob AI Enhancement Hostile Mobs (40+): Zombies, skeletons, creepers, spiders, endermen, blazes, ghasts, phantoms, guardians, pillagers, witches, wardens, and more Neutral Mobs (15+): Wolves, polar bears, bees, iron golems, llamas, pandas, dolphins, piglins, and more Passive Mobs (25+): Villagers, animals, fish - learn evasion, survival, and group tactics Boss Mobs (3): Ender Dragon, Wither, Warden - adaptive boss fights that learn from your strategies 500+ tactical behaviors: Coordinated attacks, ambush tactics, terrain usage, pack hunting, evasion patterns, and more
🌍 Living, Evolving World Hostile mobs learn combat tactics and adapt to your playstyle Passive mobs learn survival behaviors - fleeing, hiding, group defense Neutral mobs develop sophisticated hunting and defense patterns Aquatic life learns ocean navigation and predator evasion Every creature evolves based on their experiences in your world
/amai stats - View ML training progress note: Machine learning wont activate until first combat /amai info - Show mod features How It Works ALL mobs use a neural network trained via
reinforcement learning:
Observe - Mobs analyze their environment (health, nearby entities, terrain, biome, time)
Predict - Neural network calculates Q-values for each possible action
Execute - Highest-value action (90% exploitation) or random exploration (10%)
Experience - Record outcome with rewards/penalties based on survival and success
Learn - Train network using experience replay to improve decision-making
Evolve - Over time, entire ecosystems develop emergent adaptive behaviors
Mob-Specific Learning Examples
Zombies learn to coordinate group attacks and flank players
Skeletons discover optimal firing positions and kiting patterns
Creepers develop stealth approaches and explosion timing
Endermen master teleportation tactics to avoid damage
Villagers learn to recognize danger, hide effectively, and alert others
Wolves evolve pack hunting strategies
Bees coordinate swarm defense of their hives
Passive animals develop predator evasion and safe grazing routes
Guardians perfect laser focus timing and temple defense
Warden adapts sonic boom usage based on player movement patterns
**Progression **
Early world (0-100 interactions): Random exploration, learning basics
Developing world (100-500): Behavioral patterns emerge across all mob types
Mature ecosystem (500-2000): Optimized behaviors, predator-prey dynamics evolve
Advanced world (2000+): Complex emergent behaviors, mob societies develop adaptive strategies
Federation Learning Logs:
https://github.com/smokydastona/adaptive-ai-federation-logs
Development contributions from:
BossScroll