Facebook’s new AI playing robot has been a comprehensive boost to multiplayer collaboration. Media reported that the new robot could beat professional players in popular board games such as Texas Poker and Go. On Friday, the company’s AI researchers said it had overcome a challenge — creating a Hanabi card game robot that could score high on a team.
(Pictured: Wikipedia, via Cnet)
As a complex Japanese card game, it allows 2 to 5 people to play together. Each game has certain limitations and requires the participants to collaborate skillfully.
Interestingly, everyone can see the cards of other players, but they can’t see their cards. Thus, in each turn, other players can give the appropriate tips.
Players are required to arrange a set of decks in a specific order, and then add up the cards with the largest numbers in each group to get the score. In the robot game, Facebook has made some changes to its processes and strategies.
AI robots can retrieve others and make a list that can change dynamically as other robots reveal information. Because the game needs to work together to get the highest score, it is possible to use logic to reason how to play.
In the process, Facebook bots take the best action after simulating a hypothetical game. Hanabi has the potential for as many as 10 million lots, the company said.
Clearly, this AI robot can think and collaborate like a human. But in addition to card games, the study could also be used to improve applications such as self-driving cars and virtual assistants.
Facebook AI bots are understood to have scored an average of 24.61 (out of 25) in the Hanabi game in which the two men are involved.
In real life, human players have only a 60 to 75% chance of getting a perfect score, while robots can easily reach 75%.