What is OpenAI Five?
OpenAI Five is a team of five neural networks developed by OpenAI that were trained to play the popular online game, Dota 2. OpenAI Five uses reinforcement learning to learn how to play the game by playing against itself and other players. It demonstrates the capabilities of artificial intelligence in mastering complex tasks, teamwork, and strategy, and has competed against professional players, showcasing its exceptional gameplay.
How does OpenAI Five work?
OpenAI Five uses a technique called Proximal Policy Optimization (PPO) to train its neural networks. PPO is a reinforcement learning algorithm that aims to maximize the expected reward while minimizing the divergence between the old and new policies. OpenAI Five learns how to play the game by playing millions of games against itself, exploring different strategies, and updating its neural networks based on the outcomes of these games.
Key milestones of OpenAI Five:
- June 2018: OpenAI Five defeats amateur Dota 2 players in a 5v5 match.
- August 2018: OpenAI Five competes against professional Dota 2 players at The International 2018, winning one game and losing two.
- April 2019: OpenAI Five defeats the reigning world champion Dota 2 team in a best-of-three exhibition match.