With most information hidden, the game Stratego
Adding in a second neural network that guesses the identity of hidden pieces was key.
Deep Blue defeated Garry Kasparov in chess in 1997, AlphaGo triumphed over Lee Sedol in Go in 2016, and poker bots have been outperforming professionals for years. However, one traditional game, named Stratego, managed to resist being conquered by artificial intelligence. Despite DeepMind's vast budget, no machine was able to consistently beat the top human players.
A group of researchers from Carnegie Mellon, MIT, New York University, and Stanford University managed to accomplish this feat. Their AI, dubbed Ataraxos, secured a 15-1 victory over Pim Niemeijer, widely considered the best Stratego player ever, with four matches ending in a draw. This achievement came at a surprisingly low cost, requiring only 16 graphics processing units and a few thousand dollars for training.
Stratego is a game of hidden information, with each player starting with 40 military pieces, ranging from a marshal to a spy, along with bombs and a flag. Players win by capturing the opponent's flag, and their identities are only revealed when pieces collide in battle, with the weaker piece being eliminated and the stronger player's identity becoming known.
Eugene Vinitsky, a researcher at NYU and co-author of the study, highlighted the unique challenge Stratego presents, noting its extensive hidden information that unfolds over an extended period.
Written by urgent.news from Ars Technica's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
- With most information hidden, the game Stratego arstechnica.com
