Meta unveils its latest AI technology Cicero which can challenge humans in the strategy game Diplomacy.
Meta AI, which belongs to Meta platforms, aims to develop various forms of artificial intelligence by improving augmented and artificial reality technologies.
Long have humans thought that artificial intelligence would not be able to mimic their ability to create arguments. However, it seems like this limit has finally been surpassed. Meta AI researchers recently announced the creation of Cicero, the first AI to display human-level performance in Diplomacy.
How Efficient is Meta AI Cicero at Diplomacy?
According to Engadget, Meta AI has trained Cicero on 2.7 billion parameters over the course of 50,000 games of the online diplomacy game at webDiplomacy.net.
Meta AI's tests resulted in Cicero finishing in an impressive second place out of 19 competitors in a 5-game league competition, even doubling its score over its rivals' average score.
Cicero can reportedly negotiate, give advice, share information, make agreements with other players, and also figure it if one is bluffing. Additionally, Meta alleges that the AI is so adept that they personally preferred working with Cicero over Diplomacy compared to human participants.
"CICERO marks the beginning of a new era for AI that can collaborate with people in gameplay using strategic reasoning and natural language processing, and the learnings from technology like this could one day lead to intelligent assistants that can collaborate with people." according to Meta's press release.
However, Meta clarifies that Cicero is still far from perfect and occasionally still produces inconsistent language while still excellent enough to present a significant challenge to even the top Diplomacy players.
Meta also notes that perhaps the technology used by Cicero can be utilized through different means, such as a voice assistant, allowing AIs to converse with humans for an extended period of time and impart knowledge. Another possibility is the deployment of AI to create realistic characters in video games that are difficult to distinguish from real gamers.
The Development of Meta AI's Cicero
PC Mag reports that strategic reasoning and natural language processing are two of the distinct fields of AI research that were combined to create Cicero, thus, is a combination of an integrated strategic reasoning engine and a configurable discourse model.
AI technologies have long been capable of beating even the best players at games such as chess, poker, and Go using self-play reinforcement learning. However, since Diplomacy requires cooperation with human players, that method doesn't work. For these games, supervised learning is typically applied, but Meta claims that this results in an opponent that is "very weak and highly vulnerable." Thus, Meta researchers used an iterative planning algorithm that "balances dialogue consistency with rationality."
As a result, CICERO can anticipate not just the plays that other players will likely make but also the actions that those players believe CICERO will make through strategic reasoning. In order to "negotiate, offer advice, share information, and create agreements," CICERO then bases its answer with other players on a set of carefully selected strategies, natural language processing.
Related Article: Chinese Metaverse Company Appoints AI Humanoid Robot as CEO
This article is owned by TechTimes
Written by: Andi C.
