Here's a bunch of experimental self learning AIs
- You can find details and the actual live demos (play them in your browser) of my experiments directly here
- This is a simple AI that does it's best to learn how to dodge projectiles that randomly come towards it.
- This simulation will spawn 100 random AIs, each AI is randomly generated, and will receive a score based on how well it is avoiding the projectiles. The best of these AIs will be bred to hopefully create a new generation that is actually able to dodge. The longer you watch, the better the AI should become!
- This AI tries to learn how to play Flappy Bird based on trial and error.
- This simulation will spawn 25 random AIs, each AI is randomly generated, and will receive a score based on how long it stays alive. The best of these AIs will be bred to hopefully create a new generation that is actually able to dodge. The longer you watch, the better the AI should become!
- This AI tries to learn how to play Ouchy Bird.
- We take the approach of creating two seperate brains -- one brain to solve the bird being dropped to the left, and one to solve the bird being dropped to the right.
- This AI is a self learning Ninja! Throw ninja stars and the AI will learn how to dodge them!
- Move your cursor to the left hand side of the screen so automatically throw ninja stars, or move your cursor to the right hand side and click to manually throw stars.
- We're using neutral networks that learn over time to generate AIs. It will learn how to dodge based on how you attack it. How many generations until it can beat you?
- This is an AI where you need to teach it how to follow your cursor.
- Move your cursor around the screen and the AI will get points based on how close it gets to your cursor.
- Left click to restart the simulation. Can you train the perfect AI?
