Researchers have developed an AI agent that can achieve a high score on the classic video game Montezuma's Revenge after watching it played only once. The agent uses a simple algorithm to learn from the demonstration, playing through multiple levels and optimizing its performance using a reinforcement learning technique called PPO. This approach allows the agent to surpass previous results, achieving a score of 74,500. The research was conducted by OpenAI, but no individual author is credited. --- Why it matters: This matters because it demonstrates an improvement in single-demonstration learning, which could have implications for AI development and game playing. It also showcases the capabilities of PPO as a reinforcement learning algorithm. Source: https://openai.com/index/learning-montezumas-revenge-from-a-single-demonstration