A new achievement: An AI-powered robot defeats elite table tennis players.
An AI-powered robot has managed to defeat elite table tennis players, a significant achievement for a machine facing professional human athletes in a realistic competitive sport. The "Ace" robotic system, developed by Sony AI, won three out of five matches against elite players, but lost the two matches it played against professionals, and won only one round out of seven matches. This achievement is considered a milestone in the field of robotics, which has long considered table tennis – with its rapid reactions, perception and superior skill – as one of the toughest tests of the progress of this technology, according to a report by The Guardian, which was reviewed by Al Arabiya Business. The term "elite players" refers to players with a high level of skill who are ranked among the best at the local or international level, without necessarily implying that they practice the sport as a profession. In contrast, "professional players" refers to those who make a living from the sport and regularly participate in official tournaments, thus making professionalism linked to career status, while "elite" reflects performance level.
How did the robot perform?
During the matches, which were held according to the official rules of the competition, the "Ace" robot demonstrated an outstanding ability to control the ball's spin, and dealt with difficult situations such as balls touching the net. It also executed a fast reverse spin shot that one of the professionals thought was impossible. A research paper about the robot was published in the journal Nature on Wednesday, but the scientists working on the project noted that "Ace" has evolved since the paper's publication. Peter Dörr, director of Sony AI in Zurich and leader of the "Ace" project, said, "We've faced increasingly stronger opponents, and we've been able to defeat increasingly stronger opponents." Artificial intelligence researchers use a variety of games, from chess to poker and breakout, to teach programs how to make decisions in complex situations. Building an intelligent robot raises the bar to new heights, requiring the machine to effectively implement decisions in real-world scenarios. "Ace" overcomes some of the complexities associated with table tennis by having an eight-joint arm mounted on a movable base, eliminating the need to stand on two feet. Instead of relying on two eyes to see the ball, it uses a system that uses several cameras to capture images from different angles of the entire table, to track the ball's position and rotation. By magnifying the ball's logo, the camera system can estimate its rotation and axis of movement within fractions of a second it takes to reach the side of the table belonging to "Ace". The robot's ability to handle rotation and choose appropriate shots was developed through 3,000 hours of gameplay within a computer simulation, while other skills, such as serving, were derived from the techniques of professional players. Ace was not skilled at table tennis from the beginning, as in his early stages he had difficulty dealing with slow balls with weak spin, as he would return them poorly and be penalized as a result. But the robot showed outstanding performance in handling difficult balls, such as those that touch the net, as it responded very quickly to changes in their trajectory.
Players' reaction
"When I used a complicated spin serve, Ace returned a complicated spin as well, which made it difficult for me," says Rui Takenaka, an elite player. "But when I used a simple serve—what we call a no-spin serve—Ace returned a simpler serve, which made it easier for me to attack on the third shot, and I think that was the main reason I was able to win." When “Ace” executed an unusual shot, intercepting the ball early and giving it a reverse spin, former Olympic table tennis player Kenjiro Nakamura said he did not believe such a shot was possible, but now believes humans can learn it. One of the difficulties of playing against "Ace" is that the robot has no eyes to look at, no body language to read, and is not affected by pressure when the score is tied 10-10. Dor said: "Players want to see their opponents' eyes. Ace's eyes are scattered around the field and show no intent or feeling." Jan Peters, a professor of autonomous intelligent systems at Darmstadt University of Technology in Germany, has been working on developing table tennis robots. He described the project as "truly impressive," but noted that research in table tennis will not solve some of the major challenges in robotics, such as object manipulation. Peters added: "To be useful to the public, we need a lot of good traditional engineering," and said: "There will be a moment in the next decade that will change the world as much as ChatGBT did in 2022. That moment may be closer now than 2036."
0 Comments