Meta’s game-playing AI can make and break alliances like a human

Learning to play Diplomacy is a lot of work. The game involves multiple players, who make moves at the same time, but each turn is preceded by a brief negotiation in which players chat in pairs in an attempt to form alliances. After this round of negotiation, players make a decision about what to do next.

Cicero models how the other players are likely to act based on their conversations with them. It is able to figure out how players can work together to achieve their goals.

A reinforcement learning model and a large language model are used to build Cicero.

Cicero isn't flawless. Some of the messages it sent contained errors or made strategic mistakes. Humans collaborated with it over other players, according to Meta.

And it’s significant because while games like chess or Go end with a winner and a loser, real-world problems typically do not have such straightforward resolutions. Finding trade-offs and workarounds is often more valuable than winning. Meta claims that Cicero is a step toward AI that can help with a range of complex problems requiring compromise, from planning routes around busy traffic to negotiating contracts.