Make sure you watch the pre-game interview with the AlphaGo engineer. There were a few interesting bits: A) A pure neural approach, with zero lookahead, can achieve amateur dan status, and B) they are considering an AlphaGo trained against self play but no book of human games. It would be fascinating to see what it would come up with.
1 comment
[ 4.1 ms ] story [ 17.8 ms ] thread