Artificial intelligence has outstripped humanity in human games
Artificial intelligence systems reached a new level in 2017. Those systems were able to beat people several times in complex intellectual games.
Google has developed an algorithm for playing AlphaGo, which defeats the most powerful and advanced players in the world and knows how to develop a unique game strategy on its own. At the same time, the innovative AlphaGo Zero system does not even need to analyze the games played by people – it learns itself.
In December of the same year, Google developers created a universal version of the AI-based algorithm – Alpha Zero. What’s really great about it is that it does not specialize in one particular game. It is enough for a neural network to get the rules of the game, and already in the course of training it develops tactics itself. Thus, Alpha Zero independently mastered the game of chess, Go and shogi, and has already managed to beat the best algorithms in each of the games. Moreover, the neural network was not specially trained to win, but only provided a standard set of rules, and the system played with itself.
As a result, one algorithm was able to learn different games for the first time in the history of neural networks development and achieve certain success in this matter:
• In 4 hours of independent study, Alpha Zero learned how to play classical chess so well that it was able to beat the reigning champion among the programs.
• Only two hours of online learning was enough for the algorithm to beat one of the best smart systems in this segment – Elmo.
• After 8 hours of learning to play Go, Alpha Zero was able to defeat the AlphaGo program, which, by the way, beat the absolute champion Ke Jie this spring (this is a man!).
But that is not all. OpenAI created a bot that defeated a professional Dota 2 player at The International esports tournament. And the Libratus algorithm was able to beat the leading poker players during the international Texas Hold’em tournament. In the course of games, the championship will soon be behind smart neural networks, and people will have to come to terms with this.