Are we amazed or concerned by AlphaGo’s 4-1 win over Go grandmaster Lee Se-Dol? Two neural networks operating in tandem give AlphaGo the edge.The first network reduces the effective depth of the search by estimating how likely a given board position will lead to a win without chasing down every node of the search tree. The second reduces the breadth of the game, limiting the number of moves for a particular board position by learning to chose the best moves for that position. So the first network generates possible moves that the second judges their likelihood to beat the opponent.
Personally, I will only be impressed when it beats Barry Cryer at Mornington Crescent.