A cool 3D interactive demo of neuroevolution with supermario (github.com)

A new interactive 3D demo showcases the concept of neuroevolution through a Mario-style platformer, allowing neural networks to learn how to play purely via evolutionary principles. Users can access the demo effortlessly through an HTML file, experiencing a game with procedurally generated levels, physics, and various challenges. The underlying architecture features a feed-forward neural network that processes inputs related to player's surroundings and outputs game actions, utilizing a unique evolutionary algorithm that incorporates crossover, mutation, and tournament selection to improve the neural networks over generations. This project is significant for the AI/ML community as it demonstrates a practical application of neuroevolution in a visually engaging manner, illustrating how neural networks can adapt and learn through genetic principles. With real-time visualizations of the networks' activation states and functionalities, users can observe how specific inputs lead to actionable outputs while competing against AI-generated champions. The demo not only serves as an educational tool but also paves the way for further exploration in evolutionary strategies within AI, highlighting the balance between creative gameplay design and advanced machine learning techniques.