No teacher. No labels. Just points for good moves and pain for bad ones.
Idle
🐶
Woof! I'm Robo-Rover. I have never seen this maze before in my life.
📦 Step 1 — The Zero-Knowledge Start
Robo-Rover has zero instructions and zero map. Watch what pure, clueless trial-and-error actually looks like.
🏋️ Step 2 — The Scoreboard Memory
Robo-Rover now runs the maze hundreds of times in the background, remembering every reward and punishment. Watch snapshots of its progress.
⚡ Speed: 10x
🚀 Step 3 — The Flawless Run
Speed back to normal. Let's watch the fully-trained Robo-Rover use everything it learned — no more randomness, just its best strategy.
💡 Summary — The Full Learning Trilogy
Super-AI, the Island, and Robo-Rover — three completely different ways to learn.
🏫 Supervised
Learns from a teacher's answer key.
"This picture is a mango."
🏝️ Unsupervised
Learns by finding hidden patterns alone.
"These fruits look similar."
🐶 Reinforcement
Learns from rewards and punishments.
"Eating this mango wins me the game!"
🎉
Nobody showed me a single answer key. Nobody grouped anything for me. I just lived through hundreds of attempts, and the points themselves taught me how to win.