AI Has Finally Learned To Play Stratego
Researchers from Carnegie Mellon, MIT, NYU, and Stanford have built an AI called Ataraxos that finally cracked Stratego, beating four-time world champion Pim Niemeijer 15 games to one with four draws. Its key advantage was a second neural network that estimates the identities of hidden enemy pieces, letting the system search plausible game states rather than brute-force a massive…
Researchers from Carnegie Mellon, MIT, NYU, and Stanford have developed an AI called Ataraxos capable of defeating the popular board game Stratego. Ataraxos outperformed four-time world champion Pim Niemeijer in a 15-1 victory, with four games ending in draws. The key to Ataraxos' success was the integration of a second neural network, known as a belief model, which estimated the positions of hidden enemy pieces.
This innovation allowed the AI to focus on plausible game states rather than exploring the vast hidden-information space, which DeepNash, developed by DeepMind, was unable to manage due to its large search space. Ataraxos learned through self-play, engaging in 163 million games to reinforce winning moves while minimizing losses.
The algorithm adjusted strategies early in the training phase with significant changes and later with smaller, gradual modifications. This approach, combined with the use of a belief model, allowed Ataraxos to outperform DeepNash, which required extensive computational resources and training time. The breakthrough findings have been published in the journal Nature.
Written by urgent.news from Slashdot's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
