The bounding part is very cool (I caught myself leaning forward in surprise [and mild horror?] as the machine jumped over the box instead of maneuvering around it). Besides not requiring fuel, does anyone know how this is different/better/unique than Google/BostonDynamic's WildCage? (https://www.youtube.com/watch?v=wE3fmFTtP9g) I certainly agree that more people working on this is a good thing, but I'd like to understand the progress being made.
I just can't help but get the feeling that those robots like this, like Boston Dynamics dogs and their cheetah, especially ASIMO and others resembling those, are mere "medieval automatas" (http://en.wikipedia.org/wiki/Automaton#Medieval_automata) of our age. They lack something fundamental, maybe even as fundamental as shift from clock gears to digital processors.
Or you may be referring to something like the uncanny valley [1] phenomenon; you're unsatisfied with the comparison to the subject of their mimicry, in this case a cheetah. Some lasers would probably help with that.
I was actually referring to technological limitations. Just like automatas were never meant to have any kind of AI, because of limitations of manufacturing techniques, these robots are the same for advancing known techniques to the limit, but not getting any closer to the goal of cheetah-like movement in this example.
I don't know if it's clear enough, since my first post was misunderstood. It's like we have a goal to reach the moon, then mechanical automatas are Egyptian pyramids and these robots using technology and software they are using now are hot air balloons.
It doesn't really look that impressive. Sure, the tech probably is very advanced, but the gait just looks silly and quite far from the "elegance" of fast land animals. While the article does imply that "getting from bounding to galloping is not that hard", I don't think it is that easy either or they would have done it already.
exactly, animals tend to use the most energetically efficient form of locomotion for a given desired speed (walking vs. running for example). i'm not sure that bounding is more efficient than either walking or galloping at any speed, since more energy seems to go into vertical and rotational motion than those more common forms of locomotion.
I suppose people working on such projects don't think of their creations as "scary as fuck"? Because that is exactly what I think when I see an effective dog-like robot. I cannot help but imagine, for example, that in the future, campus shootings will be replaced by campus robots hunting down students and faculty (due to malicious programming)! Is the fear I feel that irrational, founded only on Hollywood movies in which mankind always ends up fighting the rise of AI?
Also, (before software and hardware eat the world, literally) I don't think the capitalistic structure + unbridled automation is a stable combination. Because those who own the capital will always prefer an uncomplaining and efficient worker, the robot, to a human being 9 out of 10 times. And those who own their labour are, well, outdated. The pace of automation is beyond unprecedented; it is incomprehensible.
I'd be more concerned about something like this robotic machine gun, which is designed to do nothing but kill humans efficiently:
http://www.ubergizmo.com/2014/09/samsung-sgr-a1-robot-sentry...
"$200,000, all weather, 5.56 mm robotic machine gun also sports an optional grenade launcher... identify and shoot a target automatically from over two miles away... robot will not distinguish between friend or foe..."
> Is the fear I feel that irrational, founded only on Hollywood movies in which mankind always ends up fighting the rise of AI?
Almost certainly, yes. AI has its dangers, but it's also potentially the source of a huge amount of innovation and the solution to a broad class of problems.
> I don't think the capitalistic structure + unbridled automation is a stable combination.
There are other structures that could potentially work in a post-scarcity world, but let's create a post-scarcity world first, and file the rest under "problems we'd love to have".
Scary? That thing is awesome. And totally not threatening. It would be able to hunt you for 10 minutes and then run out of battery. If it does catch up to you, you can always go up a stairs, or just kick it over.
AI is of course a serious concern. But this isn't AI, it's a robot.
> In labs everywhere, experimental robots would leap up from lab benches in a murderous rage, locate the door, and—with a tremendous crash—plow into it and fall over.
> Those robots lucky enough to have limbs that can operate a doorknob, or to have the door left open for them, would have to contend with deceptively tricky rubber thresholds before they could get into the hallway.
It's like looking at cars and only thinking how dangerous they might be if someone decides to run you over. And it's true, over 30,000 people die in the US every year from car accidents. But the potential seems to be worth it for most people.
capitalistic structure + unbridled automation is a stable combination
It's worth noting that the killing robots in use now are the drones in service of the United States government.
Sure, sooner or later, individuals are going to build robots that run amuck and kill some people - but to really massacre at a large scale takes a state sponsor.
Or a robot that runs amok, that cant easily be stopped by, hacking it, shooting it with rifles, shooting it with cannons, bombing it, or shooting it with missiles, cruise or otherwise.
You know... crazy mad scientist Japanese Anime Mecha or Star Wars Worldeater type stuff.
Without a doubt that is likely. But, you are missing the part where robotics will also likely significantly raise standard of living. It is also extremely likely that advanced robotics will have a further polarizing effect on wealth distribution.
In general society is moving towards significant increase in monitoring and automation, technology is making it simpler and cheaper to do so, I do not foresee reversal of this trend baring a catastrophe class event. This means that we as a society will have to seriously reexamine privacy as a whole.
one of the reasons robots like this look so unnatural is that they're trying to locomote entirely using active control (i.e., the controller tries to calculate the correct position and force continuously) with rigid limbs.
animal systems have passive-dynamic components (e.g., spring-mass-damper) that smooth out discontinuities and inherently self-corrects for many unanticipated situations. robots that incorporate passive elements (and this robot just might have some, it's hard to tell from 30 secs of video) would look more natural and have better self-stabilization.
Your first paragraph says "robots like this" then your second paragraph backpedals since you don't know if this robot has significant passive dynamics or not. Why not look it up instead of hedging?
This is very very impressive. Stable dynamic gates over even moderately bumpy terrain are Hard. Jumping is also hard and the last company to get a solid electrically powered robot to jump for bought by Google. The tricky thing though is heat. If they aren't doing active cooling then the robot will only have a few minutes of runtime. In contrast to some other commentors I think if they wanted this guy to move smoothly they could do so with this hardware, instead however they are aiming for speed.
Reminds me a lot more of a sheep than a cheetah, but still, very impressive work. It must be an incredible feeling to see something like that come to fruition.
Nicely done. I'm sure it is pretty fun to debug. All my walking robot experiments (mostly with RC servos) have had some pretty hilarious not-actually-walking moments when the gait software gets confused :-) Should be interesting to see what additional work comes out of this lab.
I would argue the Pronghorn Antelope is actually a better model to replicate since Cheetah can only maintain their speeds over short distances - as opposed to the Antelope which can maintain their 60mph speeds over much greater distances due to their larger hearts and lungs.
The technological problem of power density vs. limbs and the control of the limbs seem fairly separate to me. If we had 10 times more efficient batteries, this would simply run ten times further without any other changes. Unless the bottleneck here is heat dissipation in the motors, but even that would be way easier to fix than our battery problems.
Interesting point, but besides Cheetah being a sexier, more accessible name, the main problem they are trying to solve here is gait and top speed. The efficiency of batteries (heart/lungs) is a ancillary problem they have to solve, but not their main goal.
It's so much quieter than the gas powered Boston Dynamics robots! Bringing the volume level down is essential to many use cases, so it's good to see this work.
