Train the AI how to play pong! - Slide your mouse up and down to control your paddle. - Click to toggle data points - Aim for weak data areas (red data points) to train the AI
I coded this myself in about a day. There are more efficient methods for a pong bot than this one but it's still really cool. Javascript Version: https://greybeard42.github.io/javascript/aipong If you think this is cool, check out my Jazz improv project: https://scratch.mit.edu/projects/1047654724/ >How it works< First of all, Machine Learning is an approach to AI that records data to make future predictions. In this model, there are 3 main lists for: - Where the ball was - Where the learner was - If the learner hit the ball back The model uses this data to decide where it should move based on the most accurate data point. The learner will randomly move around the data point to add variation in the data for the future. Every time the ball hits the learner, we replace bad data or add it to the list. Every time the learner misses the ball, we save the data for the next time a ball goes in that direction.