HomeTechnologyArtificial intelligenceMeta team builds AI that plays 'Diplomacy' at very high level

Meta team builds AI that plays ‘Diplomacy’ at very high level

Science (2022). DOI: 10.1126/science.ade9097″ width=”800″ height=”530″/>
Architecture of Cicero. Cicero predicts probable human actions for each player based on board state and dialogue, and uses that as a starting point for a planning algorithm using RL-trained models. The output of planning is an action for the agent, as well as beliefs about other players’ actions, which are used to select intentions for a dialogue model to fulfill. Generated message candidates go through several filtering steps before a final message is sent. Credit: Science (2022). DOI: 10.1126/science.ade9097

A group of researchers from the Meta Fundamental AI Research Diplomacy Team (FAIR), affiliated with multiple institutions in the US, has added a negotiation component to its diplomacy-playing AI system. In a paper published in the journal Sciencedescribes the work that went into building the system and its performance when pitted against faceless human opponents.

The game or Diplomacy has been described as one of the most complex games ever devised – it recreates events surrounding World War I, when many nations were at war with many other nations, and they were not all clearly aligned. In the game, players work to negotiate the best possible outcome for the country they represent based on mutual agreements, results of battles, alignments and diplomacy.

As the name implies, one of the features of the game is diplomacy, which is a difficult skill to describe. In this new effort, the team at FAIR has improved upon an earlier version of an AI that only played the board portion of the game by adding a module that deals with real time the English language diplomacy with human players.

Must Read
How Technology Impacts Construction Efficiency

The new system is called CICERO and is built around two main components: one that processes dialogue and another that elaborates strategies. The strategy engine was essentially the same as the one in the previous version of the system. However, the team built a new engine that can be used to accept dialogue from other players, pass it on, and then receive strategic plans that are translated into dialogues sent to other players. The dialogue engine was trained using data from 50,000 conversations in games played by humans.

After building and testing the system internally, the team logged into an internet-based version of Diplomacy, where people didn’t know each other. CICERO played the game anonymously, just like all human players. So not only did it have to play the game like a human would, it had to do so without other players noticing that they were playing against a computer system.

The researchers found that only one player mistrusted CICERO’s humanity. Also, the AI ​​system proved to be a better player than most of its human opponents, scoring in the top 10% of players.

More information:
Andrew Goff et al, Playing at the human level in the game of diplomacy by combining language models with strategic reasoning, Science (2022). DOI: 10.1126/science.ade9097

Project page: ai.facebook.com/research/cicero/diplomacy/

© 2022 Science X Network

Quote: Meta team builds AI that plays ‘Diplomacy’ at very high level (2022, November 23) Retrieved November 23, 2022 from https://techxplore.com/news/2022-11-meta-team-ai-plays-diplomacy .html

This document is copyrighted. Other than fair dealing for private study or research, nothing may be reproduced without written permission. The content is provided for informational purposes only.

Must Read
The Dick-Pic-Fighting AI That Dating Apps Deserve



Please enter your comment!
Please enter your name here

Most Popular

Recent Comments