Flying under the radar of robot hype, Sony AI's Ace has shown off its rapid-speed learning abilities that are seriously remarkable, displaying powerful split-second decision-making while taking on some of the best table tennis players – and winning.
After spending a few days with all kinds of robots and humanoids recently, I've become increasingly harder to impress when it comes to their abilities. Fluid joint movements and highly dextrous fingers are perhaps the two developments that still stop me in my tracks. Sony's AI project Ace has neither of these things – but it can do something that I feel we all need to be ready for in the near future. While it could predict ball movement and easily return shots in a surprisingly offensive way, at one point it was met with something unexpected. As the human's shot saw the ball graze the top of the net and change trajectory, Ace readjusted in a split second, reacting in real time better than most humans could.
Not just that, as observers noted in the video below, Ace was able to pull out return shots more aggressively to balls that would otherwise put a player in defensive mode. While this doesn't seem like a big deal, perhaps, it's able to do something that not even the best athletes in the world could do – let alone an autonomous chunk of metal with a paddle attached to its "limb."
While we're still in the early stages of the era of self-correcting robotics, Ace's ability to make such split-second changes when the predicted ball trajectory is thrown off by clipping the net at top speed is both magical and a little scary to see.
"Artificial intelligence (AI) systems now challenge or surpass human experts in many computer games," wrote the researchers in the Nature paper detailing Ace's development. "Physical and real-time sports such as table tennis, however, remain a major open challenge because of their requirements for fast, precise and adversarial interactions near obstacles and at the edge of human reaction time. Here we present Ace, to our knowledge the first real-world autonomous system competitive with elite human table tennis players."
Of course, it didn't completely demolish the professionals it faced, but as one observer casually mentioned it seemed "super-human" in the way it played, and that wasn't hyperbole. This machine could return a ball with power and accuracy – treating an attacking shot like an easy opening to then hit back with even more force – and you don't need to be a table tennis fan to appreciate this.
If nothing else, table tennis coaches may soon be on the list of jobs that robots may soon be filling. While the video is a captivating cheat-sheet, we highly recommend reading the research detailing the development of Ace and why this particular sport was chosen for its testing.
The study was published in the journal Nature.
Source: Sony AI
