Advertisement

News

One small step for AI, One giant leap for Artificial Intelligence

Patrick Devaney

Published

AI has been having a great year this year as there has been a constant stream of impressive and innovative new uses and use cases hitting the headlines. Going beyond the text-to-image and text-to-video types of advances, however, or the impressive machine learning translation models that we’ve been seeing, is the longstanding gaming showdown that has been going on between humans and machines in games like Chess and Go.

Today, however, AI has taken a truly momentous stride as Meta has just released details about Cicero, which is an AI agent that is able to beat humans when playing the game Diplomacy. Let’s check it out and look at why this is such a big deal.

One small step for AI, One giant leap for Artificial Intelligence

Diplomacy is a popular strategy game that requires players to cooperate with each other to construct grand plans and build up or turn down alliances. It is the fact that success at Diplomacy hinges on successful human cooperation that, according to Meta at least, it has long been seen as an impossible task for AI models to compete in.

“Diplomacy has been viewed for decades as a near-impossible grand challenge in AI because it requires players to master the art of understanding other people’s motivations and perspectives; make complex plans and adjust strategies; and then use natural language to reach agreements with other people, convince them to form partnerships and alliances, and more.”

Incredibly, or terrifyingly, depending on how you look at it, Cicero is able to successfully use natural language processing to negotiate with other Diplomacy players, playing online at webDiplomacy.net. The AI model performs so well at the task that during tests it was able to score more than double the points of an average human player and ranked in the top 10 positions for participants who’d played in more than one game.

It wasn’t just Cicero’s success that was so impressive either. Meta engineers also reported that human players actually preferred allying with Cicero over allying with other humans. This means that although Cicero is only trained to play Diplomacy at the moment, in the future it could help virtual assistants hold long-term conversations. The fact that players didn’t know they were talking to a machine when dealing with Cicero marks a major step forward for AI.

In other AI news, the Linux Foundation is now in charge of running a key open-source AI framework.

You may also like