Are you smarter than a computer? No. But are you better at board games than a computer? Also, no, as a new Google AI has become a top player of strategic board games.
On the other hand, AI development does result in interesting scenarios. Case in point: Google’s DeepNash AI is now an expert in a wide selection of board games, including Stratego.
The AI platform has already been trained on a large number of board games. However, it was unable to come to grips with this specific one. Nevertheless, after years of research, the Google AI can now kick your ass at the game.
Announced on Twitter, the new AI is the first artificial intelligence program to best the strategic board game. Described as a game “more complex than Chess, Go and Poker”, this is one of the more impressive developments in virtual intelligence.
As Google explains, the player cannot directly observe the identity of their opponent’s pieces in Stratego. As such, it makes it hard for the Google AI to win against human beings consistently, but a new training method has helped to give it the edge.
The Google AI was able to learn and develop “unpredictable” strategies, beating any other bot in the game. Furthermore, it was able to maintain a “highly competitive level of play” against high-ranking human Stratego players.
Impressively, the AI program was even able to learn how to bluff its opponent, an important part of playing Stratego. These skills led the AI to gain a top 3 ranking on Gravon, a platform for expert online Stratego players.
The AI’s ability to effectively create strategies in the board game leaves the program in a unique situation to tackle complex scenarios. In a published study on the model, researchers said the model could tackle “crowd and traffic modeling, smart grid, auction design, and market problems”.
In the past few years, AI understanding of game mechanics has increased exponentially. From being able to play and understand PONG to Chess to hectic games like Overcooked, AI has massively improved in this regard.