On Thursdays, we dive deep into one of the freshest research papers or technology frameworks that is worth your attention. Our goal is to keep you up to date with new developments in AI to complement the concepts we debate in other editions of our newsletter.

Games have been at the center of the deep learning breakthroughs of the last few years. This is hardly surprising if we consider that games represent one of the most complete environments to emulate AI agents cognitive and interactive abilities. Most games used by AI agents are fundamentally based on orchestrating a series of actions in the environment but don’t involve negotiating with other players to achieve a specific outcome. Meta AI recently unveiled CICERO, an AI agent that achieved human level performance in the Diplomacy strategy game, showing the ability to negotiate and cooperate with humans.

If you never played Diplomacy you are missing a really cool experience. The game was first created in 1954 and mimics the dynamics between seven European countries in the years prior to World War I. The game involves forming alliances, territorial strategies, betrayals and many other aspects of diplomacy.

You have to been feeling pretty good to name an AI model after one of the most famous statesmen in ancient Rome. But the capabilities of CICERO are nothing to be shy of. Diplomacy has long been considered an impossible task for AI given that it requires a deep understanding of people’s motives and strategies. To enable these capabilities, CICERO relied on two fundamental areas of deep learning research: strategic reasoning and natural language processing.

Inside CICERO...