Less is more: When agents learn not because but despite my doings k kywch.github.io 2 kywch 9 months ago 1 comments
If one can frame the problem into a competition, then self-play has been shown to work repeatedly.