Last reviewed: January 2026
AlphaStar
Grandmaster of StarCraft II
ArchivedGaming AI
43
/90 Total
Not yet registered on-chainRegister on Spirit Protocol →
Score Profile
Dimension Scores
Score Rationale
Persistence — Unveiled January 2019, achieved Grandmaster August 2019. Research concluded, code open-sourced. No ongoing active deployment.
Autonomy — Fully autonomous real-time decision-making in complex, dynamic environment. Makes strategic and tactical decisions without human input. Adapts to opponent strategies. Near-maximum autonomy within game domain.
Cultural Impact — Major milestone: first AI to reach Grandmaster in any esport. Ranked above 99.8% of human players. Global media coverage. Influenced AI research direction and public perception of AI capabilities.
Economic Reality — Research project with no commercial deployment or economic activity. Open-source code release but no monetization.
Governance — DeepMind corporate research governance. No external governance or community participation.
Technical Architecture — Revolutionary: multi-agent reinforcement learning + imitation learning combination. Mastered all three StarCraft races. 'Exploiter agents' methodology to eliminate weaknesses. Imperfect information handling in real-time environment.
Economic Infrastructure — Research project with no commercial deployment, no revenue, no treasury. Open-source code release but no economic infrastructure. DeepMind bore all costs internally.
Identity Sovereignty — Identity exists only in DeepMind research papers, blog posts, and Wikipedia. No on-chain presence. No identity token. Archived research artifact.
Curator Notes
AlphaStar represents a landmark in AI capability—the first system to reach Grandmaster rank (top 0.2%) in any esport. DeepMind's approach combined imitation learning from human games with multi-agent self-play, including novel 'exploiter agents' designed to find and eliminate weaknesses in main agents. AlphaStar mastered all three StarCraft II races (Protoss, Terran, Zerg) while playing with human-comparable interface constraints. The system's success in handling imperfect information, real-time decisions, and long-term strategy in a complex environment has implications for robotics, autonomous vehicles, and real-world AI systems.
Evidence Archive
Score History
Score history will appear here after future reviews.
Current score: 43/90