How Robots Master Chores with YouTube’s Help: A Breakthrough in Machine Learning

Robots have been a part of our lives for decades, but the ability to learn and adapt like humans has been a long-standing goal in robotics. In order to thrive in unpredictable environments, robots need to do more than just respond to programming. They need to learn and adapt. A new technique has been developed that could help robots learn how to perform household tasks like opening drawers and picking up knives in just 25 minutes by watching videos of humans performing them.

The technique, called the Visual-Robotics Bridge (VRB) method, was developed by a team from Carnegie Mellon University (CMU). The VRB method allows the robot to learn how to perform the actions shown in a video, even if it is in a different setting. The robot can learn where and how humans interact with different objects through watching videos. From this knowledge, a model can be trained that enables two robots to complete similar tasks in varied environments.

The VRB method requires no human oversight and can result in new skills being learned in just 25 minutes. The robot can learn to curiously explore the world around them and be more direct with how it interacts. The more robots learn, the more they can adapt to new environments and tasks. This could enable robots to learn from the vast amount of internet and YouTube videos available.

Video is an intriguing solution that’s been the centerpiece of a lot of recent work in the space. Roughly this time last year, a CMU-developed algorithm called WHIRL (in-the-Wild Human Imitating Robot Learning) was designed to train robotic systems by watching a recording of a human executing a task. The system uses video of a human to demonstrate the task, but the update no longer requires them to execute. The computer then scans the videos thoroughly, looking for patterns while reading the subtitles for oft-used terms. It combines all these into a single sequence to produce instructions it can follow.

The VRB method is a significant step forward in the field of robotics. It could help robots learn new skills and adapt to new environments more quickly and efficiently. With the ability to learn from YouTube videos, robots could become even more versatile and useful in our daily lives.