AlphaGo レビュー

Deep Blue was the first computer program to defeat a reigning world chess champion in a match under tournament conditions.

While both Deep Blue and AlphaGo aimed to conquer complex board games, Deep Blue relied on brute-force search and extensive databases of human games, whereas AlphaGo utilized deep neural networks and reinforcement learning to develop its strategies.

AlphaZero is a generalized AI that learned to master chess, shogi, and Go from scratch, without human data or prior knowledge beyond the game rules, purely through self-play reinforcement learning.

AlphaZero represents an evolution from AlphaGo, demonstrating a more generalized and efficient learning approach by not requiring human game data for initial training, unlike the original AlphaGo. Both are DeepMind creations focused on strategic board games.

OpenAI Five mastered Dota 2, a complex real-time strategy video game that requires teamwork, coordination, and handling imperfect information, ultimately defeating world champion human teams.

Unlike AlphaGo's focus on a perfect-information board game, OpenAI Five tackled a real-time, multiplayer video game with hidden information and dynamic team play, presenting a different set of AI challenges in a collaborative environment.

Cicero achieved human-level performance in the strategy game Diplomacy, which uniquely requires natural language communication, negotiation, and the formation of alliances and deceptions.

Cicero extends beyond pure game strategy by incorporating social reasoning and natural language interaction, a dimension not present in AlphaGo's Go-playing domain, which focuses solely on board state and move prediction.