Slither.IO - BOT MAKES LEADERBOARD!

This is a video demonstration of a policy-based AI bot playing the game Slither.IO. This is a very simple, proof-of-concept, AI implementation as an exercise in learning the Slither.IO source code. A more robust AI could certainly be trained using some method of supervised learning; such efforts are on the horizon. Unfortunately, unsupervised learning appear too difficult to architect since the game relies on other human players and has to run in real time in the browser. Let me know your suggestions and feedback in the comments! I&love to hear thoughts on what input spaces could be used and strategies for effective training!