OpenAI's AI-Powered Robot Learned How To Solve a Rubik's Cube One-Handed (zdnet.com) 21
Earlier today, San Francisco-based research institute OpenAI announced that it had taught a robotic hand to solve Rubik's cube one-handed. "Lost in the shuffle is just what is new here, if anything, and what of it may or may not be machine learning and artificial intelligence -- the science in other words," writes Tiernan Ray via ZDNet. An anonymous Slashdot reader shares an excerpt from his report: The real innovation in Tuesday's announcement, from a science standpoint, is the way many versions of possible worlds were created inside the computer simulation, in an automated fashion, using an algorithm called ADR. ADR, or "Automatic domain randomization," is a way to reset the neural network at various points based on different appearances of the Rubik's cube and different positions of the robotic hand, and all kinds of physical variables, such as friction and gravity. It's done by creating thousands of variations of the values of those variables inside the computer simulator while the neural network is being trained. ADR is an algorithm that changes the variables automatically and iteratively, as the policy network is trained to solve the Rubik's cube. The ADR, in other words, is a separate piece of code that is designed to increase random variation in training data to make things increasingly hard for the policy neural network.
Using ADR, the real world Dexterous Hand can adapt to changes such as when it drops the cube on the floor and the cube is placed back in the hand at a slightly different angle. The performance of the Dexterous Hand after being trained with ADR is vastly better than without it, when only a handful (sorry again again for the pun) of random variants are thrown at it using the prior approach of manually-crafted randomness, the authors report. What's happening, they opine, is the emergence of a kind of "meta-learning." The neural network that has been trained is still, in a sense "learning" at the time it is tested on the real-world Rubik's cube. What that means is that the neural network is updating its model of what kinds of transitions can happen between states of affairs as events happen in the real world. The authors assert that they know this is happening "inside" the trained network because they see that after a perturbation -- say, the Dexterous Hand is hit with some object that interrupts its effort -- the robot's activity suddenly plunges, but then steadily improves, as if the whole policy network is adjusting to the changed state of affairs.
Using ADR, the real world Dexterous Hand can adapt to changes such as when it drops the cube on the floor and the cube is placed back in the hand at a slightly different angle. The performance of the Dexterous Hand after being trained with ADR is vastly better than without it, when only a handful (sorry again again for the pun) of random variants are thrown at it using the prior approach of manually-crafted randomness, the authors report. What's happening, they opine, is the emergence of a kind of "meta-learning." The neural network that has been trained is still, in a sense "learning" at the time it is tested on the real-world Rubik's cube. What that means is that the neural network is updating its model of what kinds of transitions can happen between states of affairs as events happen in the real world. The authors assert that they know this is happening "inside" the trained network because they see that after a perturbation -- say, the Dexterous Hand is hit with some object that interrupts its effort -- the robot's activity suddenly plunges, but then steadily improves, as if the whole policy network is adjusting to the changed state of affairs.
Re: (Score:1)
One hand? Pfft amateur! (Score:2)
Not impressed at all, someone already managed with no hand at all : https://youtu.be/xCoH2AORcEQ [youtu.be]
Re: (Score:1)
Re: (Score:1)
Re: (Score:2)
Cats are not self aware or particularly intelligent. No object permanence, no reasoning skills or understanding of cause and effect.
Yet they are very good at enslaving humans and brainwashing them into catering to their every need and want.
Cat level AI, a Rubik's Cube solver trained to give handjobs and a Real Doll body could be the end of humanity.
Re: (Score:2)
{Citation Needed}.
Whether a cat, dog, gerbil, or anything else with a brain of some size or other is 'sentient' is up for grabs, but at the current level of technology, anything with a brain has greater cognitive ability than even the best so-called 'AI', because even the best 'AI' we've written has precisely ZERO cognitive ability.
For the millionth+1 time: We don't have a
Re: (Score:2)
Does a submarine swim? I'd say so even though no fish uses a propeller. In other words, does it matter what mechanism is used if the outcomes are the same? We're not at a point where it matters or even close, but I don't think we will need a full understanding of human/animal cognition to create something that accomplishes the same tasks.
Re: (Score:2)
LOL you're either remarkably stupid, engaging in 'magical thinking' (same difference), or Just Another Stupid Troll.
In any case: just shut up, you're giving me an icecream headache.
Re: (Score:1)
Cat level AI, a Rubik's Cube solver trained to give handjobs and a Real Doll body could be the end of humanity.
Brilliant idea. Why have we not seen this yet? Reinforcement learning trained AI sexbots. Porn industry is always first to adopt the new technology (or so they say..) so... what could go wrong??? :o
Re: (Score:2)
Yeah, see the problem with that is that it doesn't really solve anything. There are literally billions of those running around and they can be had quite cheap. It's even fun to make them so people do it for free.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Humans are just chemical machines and may well be someone elses "AI" project. Moreover, their existence and the other factors I mentioned limit the applicability of the type of AI you suggest. Hardly off-topic.
Personally, I think birds are probably a better model for AI than humans. Parrots pack in a lot more intelligence per cubic mm of brain. Scaling them up to human level or above intelligence might not resemble a human at all.
"Also speak for yourself, bub."
Fair enough, to each his own. Speaking for myse
Re: (Score:2)
Oh, geez, you're one of those "human behavior is deterministic", "everything you do and say is predictable" people, aren't you?
Shoo, shoo, mister buzzkill, shoo!
Missing something important... (Score:2)
There is no actual video of it solving the cube. WTF is up with that?
Re: (Score:2)
Who cares? (Score:2)
What the important question is: Can it rub one off?
Misinterpreted (Score:2)
"Push the gurney into Exam Bay 2, stat."
"Doctor. This is the guy from the AI accident in the Dextrous Hand machine learning lab."
"Good afternoon, sir. I understand you were trying to uae the AI to train the hand to solve a problem automatically?"
"Yes."
"What happened to the skin on your wee wee?"
real world use for dextrous one handed robot? (Score:2)