Meta has announced what it calls a “breakthrough” in a specific area of AI for gaming: software called Cicero that is the first AI to achieve “human-level achievements in the hit strategy game Diplomacy”. Diplomacy is originally a board game, which has many official and unofficial digital successors, and the reason it is such an interesting choice is that the core of the game is negotiation: that is, it is a game for several players. players where the players have to constantly negotiate with each other.
The post announcing Cicero acknowledges several AI victories over humans (fact check: Deep Blue lost to Garry Kasparov before beating him several years later, after which IBM declined a rematch), but says “really useful, versatile agents will have to move on then just move pieces on a board”. So the idea is that Cicero can negotiate, persuade, and collaborate with human players to achieve strategic goals in the same way a human would.
Diplomacy has long been seen as one of the great AI challenges for exactly these reasons. You need to understand other players’ motivations, adjust strategies on the fly, and ultimately win them over. Well… Playing on webDiplomacy.net, an online version of the game, Cicero “achieved more than double the average score of the human players and was among the top 10 percent of participants who played more than one game.”
In fact, “Cicero is so effective at using natural language to negotiate with humans in diplomacy that they often preferred working with Cicero over other human participants.”
Cheating! Rank, foul betrayal!
Part of the achievement is that Cicero isn’t built on the traditional self-play reinforcement method that AIs use to learn games (playing millions of games against themselves or humans and cracking the data). Meta says it contains two main elements: “strategic reasoning, as used in agents such as AlphaGo and Pluribus, and natural language processing, as used in models such as GPT-3, BlenderBot 3, LaMDA, and OPT-175B”.
A particularly crucial part is that Cicero can recognize which players it needs to win and come up with a strategy to get them out of the way. The software “executes an iterative planning algorithm that balances dialogue consistency with rationality”, predicting players’ future moves based on dialogue before devising a plan that incorporates these predictions.
It won’t take over the world just yet: Cicero can only play Diplomacy, although of course Meta’s ambitions for this software extend far beyond an old board game. The company believes this could have a major impact on AI chat assistants, allowing them to have learning conversations and dialogues that teach people new skills, for example.
“You can also imagine a video game where the non-player characters (NPCs) can plan and converse like humans do – understand your motivations and adjust the conversation accordingly – to aid you in your quest to storm the castle.”
That’s kind of interesting: Maybe Edge magazine was right about Doom. What if you could talk to the monsters? You can read more about the technical side of Cicero and find the research report hereor see how it plays against some human experts (opens in new tab).