Competitive self-play
Researchers at OpenAI have found that simulating AIs playing against themselves can help them learn physical skills. In one example, they used this method to teach an AI how to play football, allowing it to discover skills like tackling and kicking without being explicitly programmed. This approach ensures the environment is always challenging for the AI, which improves its performance. The team believes self-play will be a key component of future powerful AI systems.
Researchers at OpenAI have found that simulating AIs playing against themselves can help them learn physical skills. In one example, they used this method to teach an AI how to play football, allowing it to discover skills like tackling and kicking without being explicitly programmed. This approach ensures the environment is always challenging for the AI, which improves its performance. The team believes self-play will be a key component of future powerful AI systems.
---
Why it matters: This matters because it shows that self-play can be used to teach AIs physical skills, potentially leading to more robust and adaptable AI systems in areas like robotics and autonomous vehicles.
Source: https://openai.com/index/competitive-self-play
This article was originally published at: https://openai.com/index/competitive-self-play