Highlights:

  • The company Meta Platforms Inc. has designed an artificial intelligence system capable of playing the popular board game “Diplomacy.”
  • Cicero consists of two parts. According to Meta, the initial component of the AI system is responsible for planning its actions throughout a game. The second component coordinates with other participants by creating natural language text.

Meta Platforms Inc. has designed an artificial intelligence system capable of playing the popular board game “Diplomacy.”

Diplomacy was formerly deemed as too complicated to play for an AI system. Two to seven players must arrange alliances with one another to win the board game. In addition, each player must attempt to anticipate the movements of their opponents.

The researchers at Meta trained their newly developed AI system – Cicero, on a dataset of 40,000 “Diplomacy” games performed by humans. The organization subsequently had Cicero play an online version of “Diplomacy.” According to Meta, the AI system’s score placed it amongst the top 10% of the contestants.

Cicero consists of two parts. According to Meta, the initial component of an AI system is responsible for planning its actions throughout a game. The second component coordinates with other participants by creating natural language text.

Meta’s researchers detailed, “CICERO is so effective at using natural language to negotiate with people in’ Diplomacy’ that they often favored working with CICERO over other human participants.”

Cicero selects its move in a given round through a multi-step procedure.

First, Cicero assesses the game’s present status and anticipates the other players’ likely movements. The AI system then utilizes this forecast to choose the actions that will optimize its likelihood of victory. From there, Cicero communicates its plans to the players it has aligned with.

Meta’s Researchers informed, “CICERO can deduce, for example, that later in the game it will need the support of one particular player and then craft a strategy to win that person’s favor – and even recognize the risks and opportunities that the player sees from their particular point of view.”

Meta has released the Cicero code under an open-source license. The business anticipates that academicians will use the code to create AI software with more sophisticated thinking skills. According to Meta, the machine-learning techniques that underpin Cicero might be implemented in many programs and applications, from chatbots to video games.

Meta’s researchers stated, “While we’ve made significant headway in this work, both the ability to robustly align language models with specific intentions and the technical (and normative) challenge of deciding on those intentions remain open and important problems. By open-sourcing the CICERO code, we hope that AI researchers can continue to build off our work responsibly.”

Cicero is the most recent, in the line of artificial intelligence systems designed to play games that are too complex for standard computer programs. Previously, researchers at Meta built the artificial intelligence system Pluribus, which can defeat human poker players. DeepMind, a subsidiary of Alphabet Inc., has constructed a neural network that has automatically learned how to play “Go,” “Chess” and other games.