Google DeepMind's robot is already playing table tennis with human players. And there's no shortage of talent – Multimedia

It may not have the level and style of a human player, but DeepMind’s bot isn’t lacking any “arms” in a game of table tennis. The bot created by Google’s AI division can now keep the ball in play and even rack up points. In the overall study, the team says the robot has already reached the level of a strong amateur human player. The system combines artificial intelligence algorithms with a robotic arm to play with a fixed racket.

Watch the video:

In the scientific document PublishedDeepMind says achieving human-level speed and performance on real-world tasks is the North Star of the robotics research community, and the new phase is a step toward that goal, with the first robot reaching the level of an amateur human player. The document highlights that “table tennis is a physically demanding sport, requiring human players to undergo years of training to reach an advanced level of proficiency.”

DeepMind's model is capable of implementing different aiming techniques and adapting to opponent attacks, which has never been seen before. There were 29 matches between humans and robots, and the robot athlete won 45% of the matches, or 13 victories. There were no known players and their skills ranged from beginner to championship level.

See pictures of the match in the gallery:

In the history of this test, he recorded one defeat against all advanced competitors, but he won all matches against beginners.It scored 55% of wins against average athletes, giving the bot a huge advantage.

In the video it is possible to see the robot's different techniques, including the ability to give the ball slight or counter effects, and adapt to the way the ball is hit. Even the fastest or slowest balls were analysed and returned accordingly. The result wasn't perfect, but it ensured that the ball continued to be in play between the two sides.

In the future, this type of robot could be used to help athletes train, with the ability to create unexpected plays and adapt to plays.This is a significant milestone, DeepMind says in the document.