That is the question...
Impressive and all...but can they make toast?
It's alive, it's terrifying, and it does perfect backflips! Boston Dynamics' gymnastic research robot Atlas has caused a minor panic on social media. With skills like this, surely humans are doomed? Here it is. Youtube Video And here's Boston's Mini Spot. Youtube Video But hold on. What could this be? It's one of Boston's …
> Given the job of moving a package of a certain weight over these obstacles, how would you design a machine to do it?
You're being too analytical. The demonstration shows what the robot can do, not that they have developed the minimum cost solution for moving a package of a certain weight.
Consider the dog: make one miniature poodle sized and give one as a 'pet' for people with Alzheimer's. The dog can then take its owner for a walk and make sure they get back home again. Program the dog with a time limit, or a distance, or whatever, and the patient can go wherever they like but can still get home.
I'll add to that.
I'm presuming that the idea behind Atlas is to learn how to do things - such as bipedal locomotion - not necessarily that Atlas is the final product.
And if Atlas is an evolutionary step to an endpoint that is Atlas-shaped- that the end is something that can fit in a human environment and achieve a multiplicity of tasks as a replacement (temporary or not) for a human- not just specifically moving a box.
Isn't the end product of things like the Darpa robot thing to come up with something that can replace an otherwise sacrificial human in a dangerous or deadly situation like "the reactor is melting down, Lassie, go into the core and fully insert the control rods or we're all goners for sure!"
Having said that, I can see where jumping could come in handy but not sure about the fancy back flips being useful. Mind if I replace that landing mat with the ball pit from a Chucky Cheese or even just a foot deep pile of styrofoam peanuts?
That first videos simply amazed me... and scared the shit it out of me at the same time.
If this kind of technology finds a "positive" usage than count me in ... However someone, somewhere is undoubtedly imagining Terminator style usage... and I for one wouldn't like to face this thing in a standoff.
This post has been deleted by its author
At the moment, I wouldn't be that worried.
Assuming it's unarmed (as even humans in most civilised countries aren't allowed to be armed), it looks like a swift kick will put it on the floor and it'll take forever to get back up. Certainly long enough for you to be long-gone.
Or even just run away. Though there are robots that "can" run fast, and robots that "can" jump up a small box, and robots that "can" do all kinds of things, and even robots sent into hostage situations to try to disarm people (always human-controlled and just used because they're expendable, not because they're actually any good) notice how they have all kinds of human operators, none of them can do them "all", and none of them really show any great skill that you couldn't observe in a police dog.
These things are really the lowest of the low on the scale of AI, because we don't really even *have* AI. Not yet. This is just a complex mechanical system with a shed-load of human-led programming and thousands of hours of very expensive testing. To jump up on a small box that's placed in a certain way. As the other demos show, likely any variation or non-marked-up obstacle will result in utter failure, not to mention actual outside human interference.
For the moment, and probably for a long time to come, "robocop" would be defeated by someone jumping on its back, shoving it into the road, or even just running past it.
For such purposes, a bi-pedal robot always was the most ridiculous idea anyway. One application of force at the right angle, and it can literally not have a leg to stand on.
I'll worry when they can change their own batteries or plug themselves in.
I move to ban lightning jacks and USB C. The inability for good old USB to be plugged in without at least 3 rotations (because old USB is actually 4 dimensional) will keep those robot servants in a cupboard beside the cheese covered toasty machine where they belong.
Early on, Google was funding them, but Google apparently saw a video from nearly 2 years ago and were NOT pleased:
apparently, after a year, they were sold to SoftBank:
Japan doesn't seem to have problems with developing robots. There's a long history of humanoid (and sometimes creepy humanoid) robots and Boston Dynamics seems to be a perfect extension of that.
I don't know why Google didn't keep them. I thought the "poor defenseless robot" video from a couple of years ago (when that guy kept teasing it like a bully might do, moving the box, shoving it on its face, etc.) was pretty good. Apparently Google didn't like it. "We at Google do NOT have a sense of humor" etc. might have been part of THAT conversation... (but I laughed, as I watched 3 stooges a lot when I was a kid, and I _do_ have a sense of humor!).
I think it was less about robots taking over, and more about NOT having a sense of humor about tripping, kicking, or pushing over a robot that's running a program telling it to pick up a box, or watching a robot almost fall on its "shiny metal ass" while walking over rough terrain in snow. [the fact that the bot recovered in an almost human-like way is pretty cool, though!].
[I'm also very glad to see these guys are still around.]
Watched the backflip video. It positively reeks of motion capture. A human might need to use its arms to counter-balance prior to and during flip execution. Would a robot really need to do the same?
Personally, I'd much rather watch the hours of outtakes where the robot spectacularly fails again... and again... and again... and...
Watched the backflip video. It positively reeks of motion capture.
I have no beef with motion capture. This isn't an AI company, it's a robotics company, and having that chunk of metal do a succesful backflip is probably somewhat difficult. I'd like to see it do a triple Salchow one day.
The arms are swinging to gain the necessary spin to do a backflip in a quick motion. You could do a flip without hands but you'd need a more powerful jump to manage the full flip with the spin generated by leaning further back before actually jumping.
> A human might need to use its arms to counter-balance prior to and during flip execution.
The arm swing is not for counterbalancing. It is for getting the rotational momentum going so that you can actually flip once your feet are off the ground. A backflip starts at your fingertips and ends at your toes.
An anthropomorphic machine would need to swing for the same reason. Especially one with a lower CoG than an actual human, such as this headless robot.
Shadow Robotics idea was quite simple.
Robots live in the human world.
Not humans live in a world made convenient for robots.
Hence a literal human skeleton (made of plywood IIRC, because it matched human levels of strength and mass better than steel) with equivalents to every muscle in the human hand (and there are lot more than the 17 joints up to the shoulder of a normal arm).
They also use a very clever, every light pneumatic muscle to keep the weight down and the response adequate (but they had trouble finding/building a noiseless 3Kw 4-8 bar air compressor, which is what a full unit needed).
AFAIK some of their work is still the SoA.
I think Atlas may be hydraulic [which is easier to manage from a technical standpoint]. It's just the sounds it makes remind me of a hydraulic system, though some of the larger motors might still be of the electric-drive variety.
pneumatics is still a good idea as well (for size and efficiency), but the hiss-hiss of the valves operating would be irritating. Hydraulic systems are a lot quieter.
So zero real-world effect.
Mainly idle idiots in US coastal areas affected.
This is also why "Russian Tweets did Hillary and Kenya" is such complete bullshit that can only emerge from the mind of a Startegy Boutique full of Generation Connect braindeads.
So the argument in this threadbare article is that, for a given very SPECIFIC problem, there is a more parsimonious mechanical solution? The global reaction to these machines is because they are a necessary sequential step towards the kinds of general purpose intelligence (and corresponding locomotion) that we possess. It's not just anthropomorphism, it's the kinds of evolutionary possibility space that were carved out by nature because they were adaptive in many different scenarios, unlike something mounted on rails or treads.