Not a game. Just dipping a toe into the murky waters of neural networks. Became interested after seeing a couple of great NN projects by @MartinBraendli and reading about the use (or potential use) of AI / NN in the law. This is a basic (and probably flawed) NN which learns to recognise just three letters (E, C and M), which vary in size and location. A letter is displayed on the screen, and all the NN knows (as its input information) is which of the 100 red/green squares the letter is touching. The NN learns by "backpropagation" based on https://stevenmiller888.github.io/mind-how-to-build-a-neural-network/. Compare its success rate at different stages, using the buttons at the bottom. NNs seem kind of magical or creepy, depending on your point of view. An NN is ultimately just a collection of inter-connected numbers, so it does feel like "black box" technology, which gives rise to transparency issues if an NN is used to make legal decisions. BTW there are a couple of great character recognition projects (using NNs) by @Miklray.
