Hey Matthew,
Turning Machine Learning and neural network design into a hands-on sandbox where humans act as the "trainers" for AI agents tackling physics challenges is a great educational concept.
Demystifying how AIs actually learn—watching them crash, adjust weights, and slowly optimize—is such a grounding counter-perspective to today's black-box hype around synthetic intelligence.
Positioning the human user as the architect/trainer of algorithmic behavior fits right into the core questions we're asking over at the Transmedia Jam #1: Humans versus AI (/jam/transmedia-jam-1).
Since pre-existing educational tools, interactive experiments, and sandboxes are 100% allowed, feel free to submit AI Toybox to the jam! It takes two clicks, costs nothing, and adds a great interactive Machine Learning perspective to the roster.
Keep up the great work! 🏎️🧠