Fish Wang
says to
YSITD
alphaGo的神經網路發現，對上koru最好的做法是rm -rf(X