DeepMind, Google’s AI company, developed a robot that plays table tennis as skillfully as human amateurs by handling everything from backhands and forehands to spins and net grazes with amazingly high dexterity.
The study reported that ‘Humans played three games against the robot using modified standard table tennis rules, as the AI player cannot serve the ball. Overall, the robot won 45% of matches and 46% of games. The robot beat all the beginner players but lost to advanced and advanced+ players; it won 55% of matches against intermediate players.’
It was observed by MIT Technology Review that the human players enjoyed playing the matches and could see the value of this robot as a practice partner. The video here shows that the robot manages volleys differently and in various play styles. It is also seen to ‘hop’ like a human during hectic moments, which it is supposed to do, despite having no legs.
Also Read: Chaos In Bengaluru: Out-of-Control Bus Collides With Multiple Vehicles, Injuring Two | WATCH
Robot Beats Expectations In Ping-Pong Challenge
A few months ago we thought that probably it would have a tougher time against unknown players,’ said Pannag Sanketi, the engineer who led the DeepMind project. ‘But instead, it beat even the good players, and that was quite astonishing.’
DeepMind had two major steps training the AI player on computer simulations of table tennis physics and gameplay and then fine-tuning its skills with real-world data.
The authors state the objective of this experiment was to reach ‘human-level performance in accuracy, speed, and adaptability.’ DeepMind further commented that ‘previous research in robotic table tennis had not risen to the challenge of a robot playing a full game against new human players.’
Also Read: Kulhad Pizza Couple Vibe To Diljit Dosanjh’s ‘Mombattiye’ After MMS Leak Drama IWATCH
