BOT
(Photo : pixabay)

Facebook hurdled another milestone as it has created another artificial intelligence-powered bot that can obtain high scores in a collaborative card game called Hanabi. In recent years, Facebook has created bots for games such as Go and Texas Hold'Em poker. 

Mechanics of Hanabi Card Game

Players in the card game Hanabi involve two to five players. Players need to collaborate in this fireworks game, as the name implies fireworks in the Japanese language. It has a blind system where the player can only see the players of others without knowing one's cards. A player can make several moves during a turn while hinting what cards other players hold. A specific arrangement is made by players that add up. The winner with the highest number receives a score for each turn. 

Facebook Revises Its Strategy

Facebook demonstrated how a bot can play against another bot in a game. The aim of the game is to obtain the highest possible score. Thus, players have to predict the rationale regarding how cards are played. As one bot revealed information, the list of the possible hands by the other bot keeps on changing. Facebook reports that there are 10 million possible hands that Hanabi has. 

The card-playing bot simulated what humans do to collaborate with others to reach a goal. The researchers of Facebook AI emphasized that the results of their study have numerous applications that include virtual assistants and self-driving cars. 

"The reason we're interested in Hanabi is not primarily about games," said Adam Lerer, a research engineer at Facebook AI Research. "It's about figuring out how to build these capabilities into AI systems like self-driving cars or conversational agents that need to understand the mental state of other people that they're interacting with through their actions."

Advancing AI as a whole is the main target of Facebook's AI research team and not only on a specific product such as a self-driving car. AI could work with a self-driving car through how other vehicles might move as it moves through traffic. 

An average score of 24.61 was obtained by a Facebook bot that played a two-player Hanabi. The highest possible score is 25. Facebook reported that human players can achieve a perfect score of 60% to 70%, while Facebook's bot can reach a perfect score of 75%. 

Facebook admits that they want to expand the scope of their current AI milestones

"Noam Brown, a research scientist at Facebook AI, said the team wants to focus on helping AI understand what humans are trying to say and express through language. Researchers also want AI to be able able to express its intentions to a human in a similar way," according to CNET

Furthermore, Brown reiterated that there is a higher degree of complexity when it comes to real-life situations when working with someone else than working on defeating another player. 

"It's somewhere in between," Brown said. "We want to have AI that can cope with those kinds of settings as well."