With most information hidden, the game Stratego
Summary
A team of researchers from Carnegie Mellon, MIT, NYU, and Stanford created an AI called Ataraxos that can beat the best human players at Stratego, a complex board game with hidden information. Ataraxos won 15 games to 1 against top player Pim Niemeijer and used a special method to guess the opponent’s hidden pieces and plan moves ahead.Key Facts
- Stratego is a game where players have 40 hidden pieces with different ranks and the goal is to capture the opponent’s flag.
- The identities of pieces are hidden until they are involved in a battle against an opposing piece.
- Stratego has a huge number of possible setups (over a decillion), and games can last around 2,000 moves.
- Ataraxos was trained by playing 163 million games against itself, learning which moves worked best.
- Unlike past AIs, Ataraxos uses a second neural network to predict the opponent’s hidden pieces and plans moves by simulating possible outcomes.
- This approach makes Ataraxos calm, careful, and strong at bluffing without being impulsive.
- Training Ataraxos required only 16 GPUs and a few thousand dollars, which is efficient compared to previous AI efforts.
- Earlier attempts like DeepMind’s DeepNash could not reliably beat top Stratego players due to the game’s complexity.
Read the Full Article
This is a fact-based summary from The Actual News. Click below to read the complete story directly from the original source.
