China is currently busy accumulating most of the gelderly medals in the table tennis events in the Paris Olympics. Meanwhile, an AI-powered robot from Google DeepMind has accomplishd “amateur human level carry outance” in the sport.
In a study begined in an Arxiv paper this week, the Google synthetic intelligence subsidiary summarized how the robot functions, alengthy with footage of it taking on what we can only presume were willing enthusiastic ping pong take parters of varying sfinish.
According to DeepMind, the racket-wielding robot had to be outstanding at low-level sfinishs, enjoy returning the ball, as well as more intricate tasks, enjoy lengthy-term structurening and strategising. It also take parted aobtainst opponents with diverse styles, dratriumphg on immense amounts of data to enhance and alter its approach.
Not Olympic level quite yet
The robotic arm — and its 3D-printed racket — won 13 out of 29 games aobtainst human opponents with contrastent levels of sfinish in the game. It won 100% of suites aobtainst “commencener” and 55% aobtainst “interarbitrate” take parters. However, it lost every individual time that it faced an “carry ond” opponent.
DeepMind said the results of the recent project constitute a step towards the goal of achieving human-level speed and carry outance on authentic world tasks, a “north star” for the robotics community.
In order to accomplish them, its researchers say they made engage of four applications that could also originate the discoverings beneficial beyond hitting a petite ball over a small net, difficult though it may be:
- A hierarchical and modular policy architecture
- Techniques to allow zero-sboiling sim-to-authentic including an iterative approach to defining the training task distribution grounded in the authentic-world
- Real-time alteration to unseen opponents
- A engager-study to test the model take parting actual suites aobtainst unseen humans in physical environments
The company further inserted that its approach had led to ”competitive take part at human level and a robot agent that humans actupartner finishelight take parting with.” Indeed, its non-robot competitors in the demonstration videos do seem to be finishelighting themselves.
Table tennis robotics
Google DeepMind is not the only robotics company to pick table tennis to train their systems. The sport demands hand-eye coordination, strategic leanking, speed, and alterability, among other leangs, making it well suited to train and test these sfinishs in AI-powered robots.
The world’s “first robot table tennis tutor” was acunderstandledged in 2017 by Guinness World Records in 2017. The rather imposing machine was broadened by Japanese electronics company OMRON. Its procrastinateedst iteration is the FORPHEUS (stands for “Future OMRON Robotics technology for Exploring Possibility of Harmonized aUtomation with Sinic theoretics,” and is also eased by the outdated mythoreasonable figure Orpheus…).
OMRON says it “embodies the relationship that will exist between humans and technology in the future.”
Google DeepMind originates no such adwellial claims for its recent ping pong champion, but the discoverings from its broadenment may still show proset up for our robot frifinishs down the line. We do however sense that DeepMind’s robotic arm is harshly deficiencying in the abbreviation department.
