Facebook, or as we shall now call them Meta, announced earlier today that their artificial intelligence CICERO has achieved “human-level performance” in the board game diplomacycharacterized by the fact that it is a game based on human interaction, not movements and maneuvers (such as chess).
Here’s a rather disturbing trailer:
If you have never played diplomacy, and you might be wondering what the big deal is, it’s a board game that was first released in the 1950s and is mostly played by people who just sit around a table (or break up into rooms) and negotiate things. There are no dice or cards that affect the game; everything is determined by people communicating with other people.
So it’s quite a bold claim for the creators of an AI to say that in a game like this it’s playing at “human level”! One that Meta supports by saying that CICERO actually works on two different planes, one cracking the game’s progress and status, the other trying to communicate with human planes in a way that we understand and with which we would interact.
Meta has brought in “diplomacy champion” Andrew Goff to back up their claims, who says, “Many human players will soften their approach or they will start to motivate themselves through revenge, and CICERO never does.” It simply plays the situation as it sees it. So it’s reckless in executing its strategy, but it’s not reckless in a way that annoys or frustrates other players.”
That sounds ideal, but as Goff says, maybe to optimal. Which reflects that while CICERO plays well enough to keep up with humans, it’s far from perfect. As Meta herself says in a blog post, CICERO “sometimes produces inconsistent dialogue that can undermine his goals,” and my own criticism would be that any example they provide of his communication (like the one below) makes him look like a psychopathic office worker who is afraid of it , if they don’t end each sentence with !!! They will think they are a terrible person.
Of course, the ultimate goal of this program isn’t to win board games. It’s easy with diplomacy as a “sandbox” for “continuing human-AI interaction”:
While CICERO can only play diplomacy, the technology behind this achievement is relevant to many real-world applications. For example, controlling natural language generation via scheduling and RL could break down communication barriers between humans and AI-powered agents. For example, today’s AI assistants excel at simple question-and-answer tasks, like telling the weather, but what if they could have a long-term conversation with the goal of teaching you a new skill? Alternatively, imagine a video game where the non-player characters (NPCs) could plot and converse like humans—understand your motivations and tailor the conversation accordingly—to aid you in your quest to storm the castle .
I may not be a billionaire Facebook exec, but instead of spending all this time and money making AI assistants better, which apparently nobody outside of AI research and corporate spending cares about, we couldn’t just… hire people who can i speak instead?
