I made a game that teaches you how AI works
Synapse is a browser game that teaches the math behind neural networks by allowing players to tune the numbers in a network to produce the right answers. The game starts with a single neuron and builds up to deep, multi-layer nets, with later levels featuring real training problems. The game includes 12 levels, a level picker, live feedback, hints, and a per-level explainer, all contained in a single self-contained HTML file.
- ▪The game teaches the math behind neural networks by allowing players to tune the numbers in a network.
- ▪The game includes 12 levels, ranging from simple to complex neural networks.
- ▪The game features live feedback, hints, and a per-level explainer to help players learn.
Opening excerpt (first ~120 words) tap to expand
Synapse_neural_net_training_game SYNAPSE is a browser game that teaches the math behind neural networks. You're given a network (inputs, weights, biases, hidden layers, and outputs) and you tune the numbers until the network produces the right answers. It starts with a single neuron (output = input × weight + bias) and builds up to deep, multi-layer nets. Later levels are real training problems: the same weights must satisfy several input→output examples at once, with decimal targets, negative weights, and genuine ReLU clamping, so guessing won't cut it, and you learn to solve them. Includes 12 levels, a level picker, live feedback, hints, and a per-level explainer. Single self-contained HTML file, no dependencies, works offline.
Excerpt limited to ~120 words for fair-use compliance. The full article is at GitHub.