For the first time, an AI-powered machine has bested elite-level athletes at a physical sport. 'Ace' is a table tennis-playing robot. It can calculate the position of a ball in space, measure its spin while it's flying over the table at 30mph and react in a tenth of the time a human player would need. Designing and training this system was no easy task...
00:00 "I'm not very good at table tennis"
00:39 Meet 'Ace'
01:09 The challenges of the sport
02:11 The history of AI vs games
04:00 Machine learning through trial and error
05:26 Systems for measuring high-speed spin
08:09 Speed, reach and reaction times
09:35 Testing the robot: The game is on!
11:09 "I really wanted our robot to win"
12:34 The potential of AI systems in the real world
Read the paper:
"
Outplaying elite table tennis players with an autonomous robot"