I can't help but think that most of the humanoid form robots floating around are and will for quite some time be more technology show cases than actual production hardware. The human form is pretty decent for a wide variety of uses, but a log more efficiency can be gained by tailoring the form for specific applications.
I'm more than a little concerned. I would think that if robots were to be used for helpful assistance to humans, they shouldn't look like they're moving in and taking over.
Now that the teams are tasked with dropping in the brains and handing them back to the Pentagon, it's even more disconcerting.
Science fiction certainly inspired this, though persuing some of these lofty goals can give us tangential gains. While researching how to properly leverage weight at the end of the arms to mantain balance, they probably learned a lot about sensor setup, programming, and even hardware. Those improvements will find themselves used in more practical projects in the future. I find it easiest to think of many of these projects along the same lines as a concept car.
With all due respect to the designers of such a technical wonder, I wonder how much the sci-fi aspect led the design compared with the functionality it would need for use in disasters. If it's sci-fi, I'd rather have it look like a frightening bug that can scale debris easily and use multiple arms for real rescue. In the absence of that, I agree with Janine that a washing machine would be a better alternative to this collectible-toy-like model.
The Form of Atlas is humanoid because DARPA has a key design assumption that human form will remain the primary design accommodation for vehicles and controls in technological environments.
This design assumption will continue until vehicles are fully automatic and humans are removed from technology control loops -- which could take either a civilization collapse -- or sudden technological disruption of current human labor political relationships through instantaneous quantum internet services.
(Humor) In its current form, Atlas could run for Congress and the Office of the President, human offices where Artificial Intelligence is not yet required.
(Humor) Atlas could takeover Jay Leno's desk without any problems and still receive incredibly high ratings for Comcast Universal NBC. AND, it wouldn't need sick time, vacation breaks, a wife, a cat, 47 different cars, or an annoying talent agent and manager. Atlas would cost peanuts for the Tonight Show to operate and it could be available 24 hours a day to support worldwide broadcast operations on the Internet in all major languages.
The fact that Atlas is still on a power tether suggests that DARPA does not have confidence in independent operations yet - beyond the test labs - since Boston Robotics has produced tetherless prototypes for projects which provide in-the-field troop support and reconnaisance.
It really isn't that hard to provide Atlas with mobile power if it was intended to be a field-ready automaton - instead of a simple test bed for autonomic bi-pedal movement and stabilization controls.
As to more appropriate robot forms for disaster response and rescue: CONSIDER THE ELEPHANT
For independent operations and ability to run on Natural Green Power, DARPA should consider issuing a Robotic Elephant Challenge, to deliver a robot which has the power and agility to not only dodge a wrecking ball - but tolerate a direct strike from a wrecking ball - AND then pick it up and throw it back.
Then again, perhaps the US Army should just consider developing an Elephant Cavalry Brigade to complement the honor guard Horse Cavalry Regiments.
Elephants have the advantage of long-distance infrasound reconnaisance and networking - and an innate ability to learn different human languages.
And, elephants use green power, consuming a minimum of 250 kg of fiber a day to generate methane gas and high value fertilizer. With instruction and management, elephants can quickly repopulate and manage forests with preferred hardwood tree crops - while selecting against invasive softwoods and weeds which only create dangerous wildfire fuels.
They can also be lawn mowers and hedge trimmers - though the boredom alone could make them irritable.
Enhanced with mobile electronics, powered by capacitive-generation fabrics, a mobile elephant platform could supply disaster response teams with needed strength, wireless networking, and intelligence to effectively extract survivors trapped in building wreckage.
The elephant's infrasound detection and imaging capability alone could help rescue teams make faster selections of survivors over dead body recovery sites.
And, there is the legendary discriminatory sense of smell which elephants have which makes them ideal for detecting dangerous gas leaks, something Atlas can not do at this time.
Given the increased risk of building damage due to the higher frequency of tornadoes and severe hurricanes, elephants and their human managers should be on every disaster response team procurement list, especially in areas with high water flooding risks.
To be blunt, there should be herds of elephants in Kansas, Oklahoma, and Louisiana, being trained and readied for emergency disaster response by the National Guard Reserves.
True, if DARPA were to issue a Robotic Elephant challenge, a one-on-one competition may still recommend the Real Elephant for practicality, maintenance economics, and general sociability.
Besides - Real Elephants are more fun, given their sense of humor, though herd ethics training will be required: Elephant teasing can be very infectious.
As always, consider the alternatives, while we still have Elephants!