Rally Between Robots... Google DeepMind Unveils Self-Learning 'Table Tennis Robot'

Google AI Model "Gemini" in Action
Competition to Secure "Physical AI" Technology

Google's artificial intelligence (AI) subsidiary, Google DeepMind, showcased a "table tennis robot" that uses two robotic arms to receive and hit a ping pong ball on the 21st (local time) through IEEE Spectrum, a scientific and technological journal published by the Institute of Electrical and Electronics Engineers (IEEE).

The newly announced table tennis robot operates by having the robots play matches against each other and learn techniques autonomously through self-play. Previously, in August of last year, DeepMind unveiled a robot capable of playing table tennis against humans.

Google's artificial intelligence (AI) subsidiary, Google DeepMind, showcased a "table tennis robot" that uses two robotic arms to receive and hit a ping pong ball on the 21st (local time) through IEEE Spectrum, a scientific and technological journal published by the Institute of Electrical and Electronics Engineers (IEEE). DeepMind video capture

The released video shows robotic arms positioned on both sides of a table tennis table, moving two-dimensionally along rails and engaging in rallies by returning balls hit by the opposing robotic arm.

This robot utilizes Google's AI model "Gemini." Gemini visually observes the robot's movements and then suggests ways to improve its performance.

DeepMind explained, "Through this table tennis robot experiment, we aim to demonstrate the potential of a robot system based on 'self-improvement,' where the robot can learn and advance autonomously without human intervention."

DeepMind's development of the AI table tennis robot is interpreted as an effort to secure technology in the field of physical AI. Physical AI refers to AI systems that learn from their surrounding environment, make autonomous decisions, and take actions accordingly, enabling physical interaction with the real world. Humanoid robots, which behave like humans, and autonomous vehicles are cited as representative industries within the physical AI sector.