Dota 2 Bot by Open AI

Dota 2 Bot by Open AI

Our Dota 2 result shows that self-play can catapult the performance of machine learning systems from far below human level to superhuman, given sufficient compute.

Conclusion is, supervised learning is ok. But if you need best performance, let your reinforcement learning system run unsupervised for a while and it’ll learn tricks that humans can’t.

Create a website or blog at WordPress.com