Business Case Studies & Business Publications

The Greatest Gamer: Reinforcement Learni...

Zorc, Saša;Hoch, J...

Technical Note

The Greatest Gamer: Reinforcement Learning

Zorc, Saša; Hoch, Julia

QA-0928 | Published June 15, 2026 | 11 Pages Technical Note

Collection: Darden School of Business

Product Details

In the 2020s, large language models have become very widespread very quickly, and they often are what people think of when they think of AI. But there is another, completely different AI that exhibits superhuman performance: game-theoretic reinforcement-learning-based (GTRL) AI. After brief introductions to game theory and reinforcement learning, this technical note offers a historical overview of the development of GTRL, beginning with its dominance in the game of Go (a "perfect-information" game, meaning all players have all the information) over the best professional human player in the world, followed by its similar dominance in poker (an "imperfect-information" game) and the real-time strategy video game StarCraft, which poses challenges similar to those in the real world. The note then delves into the current (mid-2020s) state of GTRL and its potential macroeconomic effects. Finally, the note explores potential business applications, challenges, and dangers of GTRL, including in financial markets, online security, advertising marketplaces, and geopolitics. At the University of Virginia Darden School of Business, this note is taught in the second year “Games, Cooperation, and Competition" class; it would also be suitable in a module on new developments in either game theory or artificial intelligence.