AI finally beat the best Stratego player in history—and did it on a budget
Researchers from CMU, MIT, NYU, and Stanford built Ataraxos, an AI that beat top Stratego player Pim Niemeijer 15-1 using only 16 GPUs.
An AI called Ataraxos, developed by researchers from Carnegie Mellon, MIT, NYU, and Stanford, defeated Pim Niemeijer 15 games to 1 with 4 draws in Stratego, an imperfect-information game with over a decillion possible setups and games lasting up to 2,000 moves. Training required just 16 GPUs and a few thousand dollars, contrasting with DeepMind's heavily funded DeepNash attempt from 2022. The breakthrough came from handling massive hidden information unfolding over long time scales and balancing bluffing behavior.
- Ataraxos beat top human Stratego player 15-1 with four draws
- Trained on just 16 GPUs for a few thousand dollars
- Succeeded where DeepMind's 2022 DeepNash fell short
- Stratego involves over a decillion setups and games up to 2,000 moves
From the original article
2 min read at arstechnica.com
Adding in a second neural network that guesses the identity of hidden pieces was key.
By Jacek Krywko · arstechnica.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.