General search techniques without common knowledge for imperfect-information games, and application to superhuman Fog of War chess

Researchers have developed Obscuro, the first AI system to achieve superhuman performance in Fog of War chess, a complex imperfect-information variant of chess. The breakthrough introduces new search techniques for imperfect-information games and represents the largest zero-sum game where superhuman AI performance has been demonstrated under imperfect information conditions.

• →Obscuro is the first AI to achieve superhuman performance in Fog of War chess, surpassing both previous AI systems and human world champions.
• →The system introduces novel search algorithms specifically designed for imperfect-information games that require reasoning about opponent knowledge and information gathering.
• →Fog of War chess represents the largest turn-based zero-sum game by amount of imperfect information where superhuman AI has been achieved.
• →The breakthrough extends beyond traditional chess calculation to include complex reasoning about signaling, opponent knowledge states, and information asymmetry.
• →This achievement follows the pattern of AI progress benchmarks in games, building on previous successes like poker and regular chess.