Meta announced what he calls a “breakthrough” in a certain area of gaming AI: a software called Cicero that is the first AI to achieve “human-level performance in the popular strategy game Diplomacy”. Diplomacy was originally a board game that has many official and unofficial digital successors, and the reason for such an interesting choice is that the core of the game is negotiation: that is, it is a multiplayer game in which players have to constantly bargain. each other.
The message announcing Cicero acknowledges various AI “victories” over humans (fact check: Deep Blue lost to Garry Kasparov before beating him a few years later, at which point IBM pulled out of the rematch), but said “really useful, universal agents will have to go beyond just moving pieces around the board.” As such, Cicero is designed to be able to negotiate, persuade, and work with human players to achieve strategic goals in the same way that a human would.
Diplomacy has long been considered one of the main tasks of AI for precisely these reasons. You need to understand the motives of other players, adjust strategies on the fly, and eventually win them over to your side. Well… Cicero played on webDiplomacy.net, the online version of the game, and “accomplished more than twice the average of human players and was in the top 10 percent of players who played more than one game.”
In fact: “Cicero uses natural language so effectively to negotiate with humans in diplomacy that they often preferred to work with Cicero over other human participants.”
Betrayal! Rank, heinous betrayal!
Part of the achievement is that Cicero was not built on the traditional self-reinforcing method by which AIs learn games (by playing millions of games against themselves or humans and processing data). Meta says it includes two main elements: “strategic thinking, used in agents like AlphaGo and Pluribus, and natural language processing, used in models like GPT-3, BlenderBot 3, LaMDA, and OPT-175B” .
The particularly important part is that Cicero can determine which players he needs to defeat and come up with a strategy to win them over to his side. The software “runs an iterative scheduling algorithm that balances dialogue sequencing with rationality” by predicting the players’ future actions based on the dialogue before making a plan that incorporates those predictions.
It’s not going to take over the world just yet: Cicero is only capable of playing Diplomacy, though of course the Meta’s ambitions for this software go far beyond the old board game. The company believes this could have a big impact on AI chat assistants, allowing them to have educational conversations and dialogues that teach people new skills, for example.
“Alternatively, imagine a video game where non-player characters (NPCs) could plan and talk like humans do – understanding your motives and tailoring the conversation accordingly – to aid you in your quest to storm the castle.”
Now that’s pretty interesting: maybe Edge magazine was right about Doom. What if you could talk to monsters? You can read more about the technical side of Cicero and find research paper hereor watch him play against some human experts (will open in a new tab).