> It's not clear to me how you would improve what they have now
They could improve the policy network which was based off 100,000 amateur level games. Now they could use AlphaGo self play games which are at the level of 9p as a training set.
Another thing they could do is let it run more self play games in order to improve the value net even more.
They could improve the policy network which was based off 100,000 amateur level games. Now they could use AlphaGo self play games which are at the level of 9p as a training set.
Another thing they could do is let it run more self play games in order to improve the value net even more.