I’ve always thought of top-flight AI players as a sort of death knell for strategy games. That way lies solved systems, where memorization is king. A blog post about AlphaGo, however, suggests a different use for them: exploring new tactical avenues .
In sum, the article points out that AlphaGo (a very skilled AI player of the classic game Go) sometimes does things human players have long written off as ill-advised. Never embarrassed by the risk of making a rookie mistake, and able to think about the entire board with each move, AlphaGo explores strategies that the Go world has dismissed or even actively rejected. That allows it to find not just new sequences of plays, but to reveal innovative approaches to the game.
RoboRosewater showed us that computers can help us brainstorm. AlphaGo has a new lesson on offer: if a game is deep enough, AI players can direct us toward avenues of complexity that humans have yet to explore. Computer assists can be valuable at both ends of a game’s lifecycle.