Google DeepMind Develops a 'Solidly Amateur' Table Tennis Robot (techcrunch.com) 20
An anonymous reader quotes a report from TechCrunch: In a newly published paper titled "Achieving Human Level Competitive Robot Table Tennis," Google's DeepMind Robotics team is showcasing its own work on the game. The researchers have effectively developed a "solidly amateur human-level player" when pitted against a human component. During testing, the table tennis bot was able to beat all of the beginner-level players it faced. With intermediate players, the robot won 55% of matches. It's not ready to take on pros, however. The robot lost every time it faced an advanced player. All told, the system won 45% of the 29 games it played. "This is the first robot agent capable of playing a sport with humans at human level and represents a milestone in robot learning and control," the paper claims. "However, it is also only a small step towards a long-standing goal in robotics of achieving human level performance on many useful real world skills. A lot of work remains in order to consistently achieve human-level performance on single tasks, and then beyond, in building generalist robots that are capable of performing many useful tasks, skillfully and safely interacting with humans in the real world."
The robot's biggest trouble areas are responding to fast balls, high and low balls. It also has trouble with backhand and the ability to read the spin on an incoming ball. Here's how the researchers plan to address the issue with fast balls: "To address the latency constraints that hinder the robot's reaction time to fast balls, we propose investigating advanced control algorithms and hardware optimizations. These could include exploring predictive models to anticipate ball trajectories or implementing faster communication protocols between the robot's sensors and actuators."
The robot's biggest trouble areas are responding to fast balls, high and low balls. It also has trouble with backhand and the ability to read the spin on an incoming ball. Here's how the researchers plan to address the issue with fast balls: "To address the latency constraints that hinder the robot's reaction time to fast balls, we propose investigating advanced control algorithms and hardware optimizations. These could include exploring predictive models to anticipate ball trajectories or implementing faster communication protocols between the robot's sensors and actuators."
Just a matter of time... (Score:1)
As a reasonably good chess player, I was disappointed when the programs got better than the best humans. Now I don't dwell on it much. We live in a different world, and being the best will start to mean being the best human --- in many endeavours.
Re: (Score:2)
Re: (Score:2)
In a closed system like a chess game the computer will always win. However in the real world where randomizing is introduced the human can adapt better then the computer.
--
The Global Alliance for Responsible Media is 'discontinuing' after Elon Musk's X filed an antitrust lawsuit against it [businessinsider.com]
Re: (Score:2)
Re: (Score:2)
It will if your survival is on the line. Remember you're only here because your ancestors out-ran and out-thought bigger animals trying to eat them.
Black Mirror: Metalhead [wikipedia.org]
...When Pitted Against a Human Component. (Score:2)
Re: (Score:3)
HC is the new NPC
Well, at least (Score:2)
Well, at least it's not trying to kill the competition with mercury poisoning ... yet.
Re: (Score:2)
Interesting idea (Score:2)
RL works great in a simulation environment, playing a table tennis computer game the AI would be unbeatable. But this is operating in the real world.
From my skimming of the paper it sounds like it learned the skills through simulation, but then learned how to combine them by playing in real life.
I suspect this approach could be applied to robots in general, train the fine motor skills in simulation and then teach it to combine the skills in real life.
Impressive (Score:3)
The videos are interesting. A robot arm mounted on a rail. It doesn't show the entire machine but it looks like a pretty big apparatus. Those folks in the background are apparently sitting at their daily workstations, it looks like a fun project.
https://sites.google.com/view/... [google.com]
I've done a little machine vision work and what they are doing is hard. Tracking the ball accurately and getting that arm to make accurate shots into the corners is quite an accomplishment. I'm thinking this ability to predict motions through various angles and respond appropriately in real time could be valuable in a lot of different scenarios.
"Techniques to enable zero-shot sim-to-real including an iterative approach to defining the training task distribution that is grounded in the real-world"
It sounds like they've trained up one or maybe multiple collaborative models to do the predictions.
Re: (Score:2)
https://scholar.google.com/sch... [google.com]
That's not to say this paper isn't bringing something new to the table (ha) although offhand I don't see what it is.
But it's somehow refreshing to see an article on an AI paper that's not about deep nets / transformers for the first time in a few years. Like finding your comfy old shoes in the closet.
A dystopian viewpoint (Score:2)
When I put this kind of development in the context of corporate overlords owning more and more of the world and having more and more control over the rest of us, I get really worried. If modern civilization survives long enough, it may get to the point where robots are doing all the digging, mining, building, manufacturing, playing of sports, writing of books, making of music, and so on. At that point, what purpose will exist for those of us who are not at least wealthy enough to be point-one-percenters? We
Re: (Score:2)
When I put this kind of development in the context of corporate overlords owning more and more of the world and having more and more control over the rest of us, I get really worried. If modern civilization survives long enough, it may get to the point where robots are doing all the digging, mining, building, manufacturing, playing of sports, writing of books, making of music, and so on. At that point, what purpose will exist for those of us who are not at least wealthy enough to be point-one-percenters? We may not be useful even as servants or as slaves.
Science Fiction has hinted at this, but AFAIK has never taken it to the extent of "99 percent of humanity is surplus to requirements and should die or be killed so as not to be a drain on our resources".
The idea may sound anywhere from extreme to delusionally paranoid, but I can't shake it and can't think of any strong counter-arguments. Of course, my earlier caveat of modern civilization surviving long enough to reach that stage seems to be facing some pretty long odds. So, hooray for the end of civilization, I guess?
My own theory is that we will either be ignored as we go jobless, homeless, fight each other for scraps and slowly die, or they'll line up religion with the corporate mantras in a way where, if you worship the profit, obey the rules to make more profit, and help the corporation gain profit, you may be spared, but if you make a negative impact on profit you will be tried, found guilty, and executed. At least, that's the dystopian novel I've been outlining in between writing other stories for the last ten yea
Speed Not A Robot Advantage? (Score:2)
"To address the latency constraints that hinder the robot's reaction time to fast balls, we propose investigating advanced control algorithms and hardware optimizations. These could include exploring predictive models to anticipate ball trajectories or implementing faster communication protocols between the robot's sensors and actuators."
Given the inherent speed advantage of electronics over 120 m/s nerve impulses and the capabilities of mechanical actuators over human muscle it is somewhat surprising that this is an issue. The immense parallel processing of the human nervous system may in fact be providing predictive modeling of the opponents trajectories.
Re: (Score:2)
Re: (Score:3)
I used to play in my youth.. This game often involves apparent reaction times faster than humanly (or machinely?) possible. What you're actually doing is placing the ball to limit the range of possible responses, and reading the opponent's body language to understand what their response will be (hopefully before they do). You move into position for your next shot as the ball is moving away from you, not after the opponent hits the ball. When you get in the flow, it's magical. This is going to be a very diff
Last bastion of human v. robots won't be sports. (Score:2)
Study (Score:1)