@6loss @cmhobbs !go it has all the existing databases of games with all the games played in competitions or online (all of them, including Lee Sedol's), and it has them organized, developped in decission trees, etc. Playing against AlphaGo is to play against an statistical study on the most effective move for every situation of the board ever existed. There is no «intelligence», no «strategy», no «porpouse» in AlphaGo, just an answer for almost every move. And when there is no clear answer (because there are not enough statistical evidence) the heuristic algorithm calculates using reinforced Montecarlo methods.

AlphaGo has nothing to do with IA, we learnt nothing of intelligence thanks to it. It is an exercise of prediction and quick response on behaviour. Google is paying it not because of love to Go, but as an exercise of the technologies they will need to control and detect social demands, crime or whaterver they want to find in order to make business in the future.