Why did AI researchers turn to StarCraft and Dota 2 after conquering chess and Go?
Chess is hard. But real-time strategy games with fog of war and thousands of possible actions per second? That's a whole different beast.
The answer
After mastering perfect-information board games, AI researchers tackled real-time strategy games that feature imperfect information, long-term planning, real-time decision-making, and enormous action spaces. In August 2017, OpenAI's bot defeated top professional Dota 2 players in 1v1 matches, and by April 2019, OpenAI Five beat the reigning Dota 2 world champions OG in a full 5v5 game after training for the equivalent of 45,000 years of gameplay. Meanwhile, DeepMind's AlphaStar reached Grandmaster level in StarCraft II in October 2019, placing in the top 0.2% of all human players on the European ladder. These achievements required breakthroughs in multi-agent coordination, long-horizon planning, and operating under uncertainty - skills far more relevant to real-world AI challenges.
Source: Nature (Vinyals et al., 2019)
- gaming ai
- starcraft
- dota
- reinforcement learning
More from AI & Machine Learning
- What radical question did Alan Turing ask in 1950 that launched the entire field of artificial intelligence?
- Can a machine fool you into thinking it's human - and does that mean it's truly intelligent?
- What happened at a small Dartmouth workshop in 1956 that gave artificial intelligence its name?
- How did a 1958 invention called the Perceptron convince people that machines could learn like brains?
- Could a 1966 chatbot really make people believe they were talking to a therapist?
- What was the first robot that could 'reason' about its own actions?