OpenAI researchers have used deep reinforcement learning to play the popular multiplayer game Dota 2 at a high level. The team's approach involves training an AI agent using large amounts of data from online gameplay, allowing it to learn complex strategies and tactics. According to OpenAI, their system can play against human opponents with a win rate of around 90%. The research demonstrates the potential for deep reinforcement learning in complex domains such as game playing. --- Why it matters: This matters because it shows that large-scale deep reinforcement learning can be applied to complex tasks like game playing, which could have implications for other areas such as robotics and autonomous systems. Source: https://openai.com/index/dota-2-with-large-scale-deep-reinforcement-learning