Overview

• Researchers from MIT, Harvard University, and the University of Washington have developed a new reinforcement learning approach, Human Guided Exploration (HuGE), that uses crowdsourced feedback to train AI agents.
• HuGE allows AI agents to learn more quickly, despite the fact that data crowdsourced from users are often full of errors.
• The new approach allows feedback to be gathered asynchronously, enabling nonexpert users around the world to contribute to teaching the agent.
• In real-world and simulated experiments, HuGE helped agents learn to achieve the goal faster than other methods.
• In the future, this method could help a robot learn to perform specific tasks in a user's home quickly, without the owner needing to show the robot physical examples of each task.