On the arXiv database, engineers from Google's DeepMind artificial intelligence (AI) project have just announced a remarkable achievement: A robot capable of playing table tennis at an amateur level.

The robot is developed based on the ABB IRB 1100 industrial robot arm, equipped with the ability to slide horizontally on rails.

To teach the robot to play table tennis, the researchers used a two-level approach. Level 1 focused on performing racket movements and level 2 on developing strategies.

The robot's AI system was trained by observing physical simulations and human matches. After training, the robot participated in 29 matches against players of various levels.

The results showed that the robot could easily beat novice players, winning more than half of the matches against intermediate players, but still could not beat highly skilled players.

Researchers have identified several weaknesses in the robot, such as slow reactions to fast and high balls, difficulty reading spin and the ability to hit backhands.

This project marks an important step forward in combining robotics and artificial intelligence, opening up many possibilities for future applications.