Training a PowerPoint AI to Play Tic Tac Toe
summary
Explains how a reinforcement learner is implemented inside PowerPoint using slides as states and animations as stochastic choices, inspired by MENACE matchbox idea. Builds a deck of about 1830 slides with hundreds of invisible shapes per state that randomly blink to represent move probabilities. Training runs many simulated games, adding shapes on wins and deleting them on losses; after thousands of iterations the deck reliably learns the best opening moves and often wins or forces draws.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:24What machine learning is
- 3:50Math isn’t the only path
- 7:03Iterative knowledge transfer
- 10:11Core ML steps
- 14:46Real-world ML example
- 18:21Tests and scores in vague scenarios
- 24:56ML applies to all problems
- 26:44Tic-Tac-Toe AI challenge
- 29:12PowerPoint AI inspiration
- 31:40PowerPoint state management
- 36:12Training with scores and punishment
- 41:09Training process breakdown
- 42:35Why ML fundamentals matter
- 46:06ML impact on problem-solving
- 47:29Closing and resources
related talks
16mMachine Learning with Matches and Confetti
31mDesigning game AI with reinforcement learning
06mCreating and Teaching Games in Python to Teach Artificial Intelligence Concepts
02mDesigning a Video Game AI
15mAI Self - Learning Game Playing
15mMachine learning with matches and confetti
46mAI Takeover
17mTraining My Rival in Java: A Deep Q-Learning AI to Play Azul