Summer of Math Exposition

In this video I use the Godot game engine and Python to train a convolutional neural network from scratch which reads handwriting done in game and then uses it to fun effect. Moreover, I use this as a framing device to go over the basics of how neural networks and deep networks learn, and the rationale behind their fundamentals and the reasoning behind moving from traditional neurons to convolutional layers (and how they work). I try to keep it fun and light along the way, and have some interesting tech-demo-type outputs at the end. I believe there should be no issues as far as the AI policy, as I trained the network completely from scratch, the architecture is of my own design, and the input data is the EMNIST letters, which are fully publicly available for any use.