While China is busy winning most of the gold medals in table tennis at the Paris Olympics, a robot created by Google DeepMind has reached a new level. This AI-powered robot can play table tennis at an “amateur human level.”
In a recent study published by Google DeepMind, the company shared how the robot works, including videos of it playing against human opponents with different skill levels. To play well, the robot had to master basic skills like returning the ball and more complex tasks like planning and strategy. It also learned to adapt its play style by using a large amount of data.
Not Yet an Olympic Champion
The robot, equipped with a 3D-printed racket, won 13 out of 29 games against human players of various skill levels. It won all its matches against beginners and 55% of the time against intermediate players. However, it lost every game against advanced players. DeepMind stated that this project is a step closer to reaching human-level performance in real-world tasks, which is a major goal for robotics experts.
To achieve these results, researchers used four key techniques that could have uses beyond table tennis:
- A system that breaks down tasks into smaller steps.
- A way to train the robot that closely matches real-world conditions.
- The ability to adapt in real-time to new opponents.
- Testing the robot in actual matches against human players.
DeepMind mentioned that their approach led to competitive play at a human level and created a robot that humans enjoy playing against. The human players in the demonstration videos seemed to be having fun.
Table Tennis Robots
Google DeepMind isn’t the only company using table tennis to train robots. The sport is great for developing skills like hand-eye coordination, strategic thinking, speed, and adaptability in AI-powered robots.
In 2017, Japanese electronics company OMRON developed the world’s first robot table tennis tutor, recognized by Guinness World Records. OMRON’s latest version, called FORPHEUS, symbolizes the future relationship between humans and technology.
While Google DeepMind’s robot doesn’t claim to represent the future of human-robot interaction, the technology developed for it could have significant implications for robots in the future. However, the robot could use a catchier name!
Source: thenextweb
