Communications of the ACM
Facebook's AI Masters the Card Game Hanabi
Facebook researchers taught the company's artificial intelligence software how to play a Solitaire-like card game that requires players to work together.
Credit: R&R Games
Researchers at Facebook have taught the company's artificial intelligence (AI) software how to play the game Hanabi, a Solitaire-like card game that requires players to work together.
Hanabi is viewed as a capable testbed for AI because it requires teamwork and strategy.
The Facebook researchers had to find a way to give the Hanabi bots a method to understand the hints of their teammate bots, based on the limited information they have about their own cards.
The researchers helped the bots using a variant of the Monte Carlo "search" technology technique to help them evaluate their possible moves. The modified technique uses a combination of reinforcement learning and the so-called multi-search technique, which helped the bots learn to play Hanabi on their own and in tandem with one another.
The researchers found that digital bots playing the game learned to help each other achieve higher scores than previous AI systems.
From Fortune
View Full Article
Abstracts Copyright © 2019 SmithBucklin, Washington, DC, USA
No entries found