Hacker News new | comments | show | ask | jobs | submit login
DARPA and Boston Dynamics unveil a new humanoid robot (extremetech.com)
123 points by Dj_Anthony2013 1537 days ago | hide | past | web | 89 comments | favorite

I'm getting tired of extremetech blogspam and sensationalism. This source video was posted yesterday by me and didn't really make it anywhere. It seems I should get a voting ring and embellish my titles a bit more.

If anyone wants to read up on the DRC the two latest press releases are here: http://www.darpa.mil/NewsEvents/Releases/2013/07/11.aspx http://www.darpa.mil/NewsEvents/Releases/2013/06/27.aspx.

I don't know. This pretty much looks like the Terminator to me. It just needs some good software and a gun. Google's self-driving car technology could be adapted into a formidable killing machine brain.

"It just needs some good software."

This is of course the hard part. But this robot was built to be a common platform for DARPA to build a project around: the DARPA Humanoid Challenge. The idea is to supply a very expensive robot to a few groups and a simulation of it to lots of groups, and have them compete on the control software. No guns this time around.

Well, a bipedal humanoid robot that can walk on any terrain and can't easily be pushed over was also the hard part, and those problems are starting to look solved.

Probably before these things are autonomous we're going to have these as human controlled drones, which is a pretty scary thought all by itself.

I know what you mean. If it doesn't scream READ ME it doesn't get anywhere...

However that's just how our brains work.

Unlike ATLAS... we're only Human

Thanks for the links,

I don't quite get the complaint though. If the goal was to get the most readers good information than any article that makes it to the front page can be a vehicle for contributing your links too. Lots of things change whether or not something makes it to the front page, but if the topic is interesting enough at least one of the stories will and that serves the purpose.

Terminator has always been one of my favorite movies, and reality has an uncanny way of getting us closer and closer to the fantasy. We justify this new technology in the same way - "it's for our protection"!

Dictators run into trouble when the armies decide they won't attack a local population because, hey, it's their families and friends we're talking about. Robots eliminate that pesky detail. Imagine how a secession crisis, based on, say, widespread spying by the government on its own citizens, might be dealt with today versus fifty years from now.


If you were a resident of Afghanistan, the concept of a war against the robots is already a reality. From that perspective, is a fairly minor detail that the robots are currently piloted by people and not computers. And I expect the AI would get effective pretty quickly - I'm sure they will develop some effective way to identify a potential combatant based on either holding a weapon or some body language/temperature cues.

This technology might not be field ready right now, but it won't be long. It's reasonable to expect that with some work, it would far exceed human capacity upon a few key dimensions - running speed, targeting ability, carrying armor/defensibility. Just like no human being can outrun even the crappiest car over any reasonable distance.

One day my dad spotted me playing Mech Warrior, and asked:

"Why bipedal war machines? That is stupid, just push them over and they now are useless."

After that, I never figured why someone would build a bipedal war machine, even though I still have my nerdgasms watching all the mecha anime, movies and games...

Anyone has a idea of why the hell someone would build a bipedal war machine?

That assumes the robots won't be radically superior at balancing than humans are; they will be. It also assumes you can push one over to begin with; you won't easily be able to. They'll be able to withstand far more force against them than a human can before falling over, and they'll have a nearly perfectly calibrated response to achieve rapid re-balancing.

As to whether it's ideal (versus whether it'll work), that's a very good question. Perhaps if you can actually make it work very effectively, it's 1) maybe (?) cheaper than having three legs, 2) more agile than a tank tread style and 3) easily navigates in/around and uses human things.

Clearly in the future there will be a plethora of robotic war machine styles for different purposes, for the same reason we use different weapons for different purposes.

Perhaps it's worth noting that Boston Dynamics' earlier 'BigDog' robots were not bipedal and apparently built like a robotic pack mule. Bipedal examples may have uses as human analogs for misdirection in war, field testing testing equipment such as hazmat suits/armor, or for 'live fire' training exercises.

That is stupid, just push them over and they now are useless.

That depends entirely on whether said machines can right themselves after being tipped over.

But, the way I see it, human cities and human technology are made for our bipedal forms, two legs and hands with five fingers and such. Having a robot match that form makes it easier to exploit urban terrain and implements in the environment. I mean, which would you theoretically rather have, a robot that is a vehicle, or a robot that can operate any vehicle it comes across?

>Anyone has a idea of why the hell someone would build a bipedal war machine?

Off the top of my head...

1.) I'd have think there's something to bipedal locomotion considering the evolutionary success it has had.

2.) We're bipedal. As long as pilots we'd recognize as human are involved, having an analog to the kind of movement they've necessarily trained a lifetime for seems sensible.

What are you using to define evolutionary success? Most pedal species aren't bipedal, and the more populous species don't even have legs.

Evolution does not necessarily always find the best solution. It can get stuck in local optima: http://en.wikipedia.org/wiki/Rotating_locomotion_in_living_s...

Perhaps less important for war machines (particularly giant ones): the world that we have built for ourselves is built for bipeds. (ADA required renovations notwithstanding)

If you find someone who has built a bipedal war machine, you should ask them. In the meantime, these machines were designed for disaster recovery. The author of the article is imagining what an army of them would look like on a battlefield but states pretty clearly that that's not what these are being designed for.

If I were designing robots for battle, I would probably make them something like armed jackrabbits with wings designed to attack in swarms.

No, octopedal is plainly better -- especially given people's fears of spiders.

> Anyone has a idea of why the hell someone would build a bipedal war machine?

People get confused operating other body-forms, and when strapped into a waldo style remote rig they need something close to human for ease of use?

Pretty much all combat soldiers in history have been bipedal. "Just push them over" is not a great plan against the US Marine Corps, and probably wasn't great against a Roman legion.

Hands? Being able to use tools like us humans would clearly be an advantage.

Because it's creepiest?

I think this all hinges on when or if these things can get produced/maintained for less than or close to the cost of a human soldier. Imagine if these got down to $10,000 to produce. Then it would be cheaper to arm a 1-2MM unit militia of humanoid robots than it would actual human marines.

Human soldiers are exceptionally expensive. You can see that in both traditional foot-soldiers and in other areas like ships and planes, where the costs skyrocket because of needing to design around / for the safety of people first.

Robot soldiers will be far less expensive than human soldiers in every respect 50 years from now. A robot soldier can be powered down, and stop consuming resources when not needed (which is most of the time).

If you could produce a robot soldier that did what humans do, with a 5 year shelf life, and you could sell it for $400k to $500k, they'd sell like crazy. Just match the 5 year cost of a human soldier, then by removing the human soldier casualty angle it becomes well worth it to a government.

The value of a combat robot is in avoiding casualties rather than cost savings. Casualties lose America wars, because it is a democracy, and it fights without an existential threat.

However, I think your numbers for the cost of humans are way low. The US army put the cost to train a soldier at $150'000. In addition, there is pay, and veterans benefits. The US will spend 1.5 trillion on veterans between now and 2022.

Isn't this a nuclear weapon vs nuclear power type issue? We'll always have the capacity for evil, I'd expect an international ban on the use of robotic soliders after some period of mass destruction (we'll still have them sitting in a bunker just in case)

I don't think so. I don't see them being as destructive as that. I think by the time they get to a level that's bannable, there will be far greater warfare threats, like millions of inch-long artificially intelligent flying swarm syringes(OF DEATH) filled with arsenic.

I think they'll be far more useful as support units.

Good point, These Darpa terminators are just the visible threat, its the classified nano robots that control us from inside that we should fear

I think you're joking, but what you just said would make a great parody of the nuts over in that other thread that are hypothesizing that the NSA is possibly keeping 5D glass hard drive technology a secret from us.

classified nanorobot technology is no joke.

I certainly hope not! They're going to need it when Cyber-Cthulhu gets loose.

The difference is that nuclear weapons are indiscriminate, whereas robotic soldiers can theoretically limit targets only to evildoers (or dissidents, undesirables, etc).

The stuff that Boston Dynamics is doing with DARPA funding feels like a current-generation parallel to how DARPA funded the early efforts at building the Internet.

I suspect we will look back on these guys as some of the early pioneers of the yet-to-come ubiquitous robotics era.

ARPA funded the early efforts of building the internet.... the D came later

Thanks for the clarification

Both, really: 'ARPA, was renamed to "DARPA" (for Defense) in March 1972, then renamed "ARPA" again in February 1993, and then renamed "DARPA" again in March 1996.' (Copied from wikipedia.)

And each time some General somewhere noted 'Reorganized unit to align with modern mission goals' as a performance evaluation bullet item for themselves.

God this is awful. Just a matter of time before these beasts are used to serve search warrants, deal with "difficult" protests, etc.

    [Dick Jones directs Kinney to threaten ED-209. Kinney points a gun at the robot.]
    ED-209: Please put down your weapon. You have 20 seconds to comply.
    Dick Jones: I think you'd better do what he says, Mr. Kinney.
    [Alarmed, Kinney quickly tosses the gun away. ED-209 steps forward and growls menacingly.]
    ED-209: You now have 15 seconds to comply. You are in direct violation of Penal Code 1.13, Section 9.
    [Everyone in the room panics; Kinney tries to hide among them, but is pushed back into open range]
    ED-209: You have 5 seconds to comply. 4. 3. 2. 1. I am now authorized to use physical force.

Or on the flip side, pulling someone from a burning building or performing CPR.

Judging by the budgets of the army and law enforcement against the various wars, abroad and local (drug enforcement for instance), I doubt that a lot of these robots will find their way in the underfunded fire stations or emergency services...

Furthermore, since defence agencies are financing the development, it's more likely they will program the robots to be sharp shooters before they learn the more complex tasks of pulling people safely out of rubble and giving them CPR.

When the US uses robots like this in war (and it will) we will enjoy a significant reduction in casualties and probably an increase in battle victories. This will probably embolden future US leaders, similar to when the US had the only nuclear weapons.

During the cold war we learned the hard way that time lowers costs of technology, and soon everyone had nuclear weapons.

I am sure the same will happen with robotic warriors and I fear the day when a terrorist can pilot a drone over my home and kill me with the same prejudice we use when piloting drones in the middle east today.

My only hope is that some politicians also understand this and move to create international robotic warrior regulations and treaties. I am skeptical of this happening however. It is always more fun to lord your power over your enemies today than to have some foresight.

Having an international inspection system to watch for the development of nuclear weapons would seem a much easier task than one that would be needed to monitor for the development of robotic warrior/weapon systems. The pieces needed for the later can much more easily look like normal production/assembly tool development given that they are, in fact, largely the same pieces.

Used to think these things were cool but now that I can't trust my government anymore... I find them deeply disturbing.

Just because they are being transparent with their research?

50 years ago the US govt had amazingly cool technology (SR-71, satellites, nukes, computers, submarines, lasers, chemical weapons) but it was kept halfway secretive.

Just because some people are uncomfortable with new technology shouldn't mean that the government is untrustworthy or evil. If anything, the transparency makes them more trustworthy.

The stuff the CDC works with is 1,000 times more destructive than any humanoid robot prototype. Imagine if they posted a video, "well, this little vial of liquid will cause your skin to fall off and your organs to melt."

Do you think other countries are going to avoid this type of research? Humanoid robots have been developed in Japan for 20 years.

I think a video tour of the CDC's vaults would be quite interesting, personally.

What is the point of having it be bipedal?

Would not the dog/four legged system work out better? You could still have it manipulate items with forelegs if you wanted but keep the inherit stability

Most robots are not bipedal, for the reasons you state, among others. But if we want humanoid robots ever, we have to get started working with the form factor. There are plenty of technical problems to solve that are specific to this kind of platform.

This robot => (more steps here ) => C3-PO => Terminator => Roy Batty.

DARPA has a great record of pushing hard on technology that might sorta kinda work, but if it does it will have huge impact. Some of it works out better and faster than expected, with the ARPAnet as the prime example.

Because most of the parts of the world that are designed, are designed for bipedal things that occupy about 1.5 feet^2, and are about 6ft tall.

Think about bigdog trying to walk through a house. It would have a lot of problems, because the house wasn't designed for something that looks like that.

Well, actually big dogs do quite well walking through a house. They're quite popular. They just need to make BigDog a little smaller, and I think the four-legs approach would work quite well.

The reason is that human size and shape allows for replacing a human in any situation such as driving a vehicle or fitting into airline seats or what have you.

A robot that drives a vehicle can be an integral part of that vehicle, likely far cheaper than a humanoid robot that fits in the seat (and also shares the human driver's disadvantage of poor peripheral vision.)

Then how about being able to be a passenger in a vehicle with any special modification?

Exactly. Entering a vehicle was one of the challenges in the most recent DARPA evaluation of teams developing high level control software for this robot.

You could just use a pickup truck.

Doors, stairs, and ladders.

I don't think 4 legs would be better.

As you can see, balance is not an issue with quick enough joints and good software. The current ATLAS can easily balance on one leg. The other leg is just there to provide locomotion.

One of the original stated goals of Petman was to test uniform designs.

Disaster recovery, assistance and space applications (asteroid mining?) are the only real applications of these robots.

This is not going to be useful in the military. The only advantage it may have over humans is increased accuracy, and even then, a mobile gun platform on wheels will do a much better job.

This robot is less mobile, less reliable, less endurable, much less smart and much less flexible than any average soldier. The only applications I can think of are extreme weather conditions (i.e. Antarctica and the Sahara desert), and even then it would probably fail fast.

I don't think it's complete garbage, it's just that I don't see it being close to useful in real military applications this century.

There are many applications of humanoid robots when they become good and cheap. Right now they are neither, but DARPA is funding people to work on this.

Cheap, good humanoid robots will fill the slave niche in society, without the ethical problems (at least in terms of the slaves themselves - it might not be good for us to be slave owners, but we'll see how that pans out). Who said the researchers are or should be thinking about only this century?

By the way, a mobile gun platform on wheels is not going to be great in built-up areas. This is why we don't have combat soldiers fighting in powered wheelchairs, and why Daleks suck.

> ...without the ethical problems...

I think that is a bit too optimistic. Political battles over robots' rights are inevitable.

Edit: I think this is also relevant: http://www-formal.stanford.edu/jmc/robotandbaby/

OK, how about "greatly reduced ethical problems".

As long as we're still killing and eating animals, we can make robots work for us.

Also, ideally we'll build them to like it. Then it'd be unkind to let them rest.

"Real" military applications, perhaps not.

"Pseudo-military" operations, i.e. in our rapidly militarizing civilian police forces, perhaps not so far off.

You think a police robot is easier than a soldier robot? Why?

Both services are deploying robots as fast as they can. Thousands are in Iraq and Afghanistan.

Cost of training per role is probably a pretty good metric.

The cost of training a human pilot is higher than a shoe-store clerk. Software already pilots planes very well, but working in a shoe store is very hard for a robot.

> Thousands are in Iraq and Afghanistan

Those are not humanoid.

Right, but the grandparent suggested that we'd see such robots in police applications before military. I'm asking why he thinks this is so. We are not seeing this with the non-humanoids.

You are raising a good point, but you are probably wrong ("this century").

Tracked vehicles run by joystick are now used operationally for IEDs. There are resupply and "lug extra cargo" missions that Big Dog is very close to fulfilling right now. There are "augment human teams" missions that smaller legged, tracked, and wheeled platforms are getting close to fulfilling. (Go into this building and look around, take up this position and guard this building entrance while the rest of the team goes in elsewhere.)

The humanoid robot can perform these tasks in a more versatile way, as mentioned elsewhere in these comments, because it can adapt to many real world situations where humans are present (e.g., doors, stairs, ladders, vehicles).

Now all we have to do is scale it up and we'll be ready for when the kaiju attack.

Correction: now all we have to do is extrude drills from it and we'll be ready for when the Anti-Spirals attack.


So pumped to see that!

Congratulations on making me snort tea up my nose :)

I'd like to know about its energy consumption, and how long would it operate with today's best battery technologies. That seems to me to be one of the bigger hurdles for turning this into something operational, but I might be wrong. Still, it's 330 lbs to move on two legs (future versions of course will be lighter, but, being military, it probably can't be all lightweight stuff). Anyone here with more knowledge about this care to give their opinion?

Slightly off-topic, but the video of this thing's predecessor, BigDog, is simultaneously fascinating and hilarious: https://www.youtube.com/watch?feature=player_embedded&v=2jvL...

The current geek want it, the future geek with the robotic gun in his head don't.

Considering how hard it is for bi-pedal bots to maintain balance, you could just kick it over and walk away.

This one is quite good at maintaining balance. It is demonstrated in the video.

What's the advantage of making the robot humanoid?

Doorways, access hatches, and the like are designed for human-sized and human-shaped entities to pass through them. Stairs are problematic for wheeled locomotion.

Interacting with a world built for humanoids, I would guess.

Looks like it has the Matrix of Leadership in its chest, just like Optimus Prime.

Damn, no robot balls to kick. There goes my plan.

I think this is incredible actually. It's worth noting they tend to try not to show power tether feeding the thing. But once they get it all nailed it's gonna be amazing. Truly awesome.

Why does it have two lenses on its head instead of just one?

Two video cameras allow stereo vision for placing obstacles, etc., into a map for motion planning.

This robot also uses a laser range finder for mapping, but there are trade offs between the two approaches, so often they are used together.

We really need to ratify and institute Isaac Asimov's 3 Laws of Robotics (https://en.wikipedia.org/wiki/Three_Laws_of_Robotics) otherwise I'm afraid our future selves will regret it.

The three laws worked really well as a literary device exactly because they are completely inadequate as an design / engineering / ethical tool. The stories are about the problems caused by the three laws, and how interesting they are. No one who has thought about it a bit, and certainly not Asimov, would seriously recommend them as an engineering solution.

Moreover, he sidestepped misuse of the tech by claiming that the Laws were "hard-wired" into the design of the positronic brain. Not only could be the Laws not be over-written, it was literally impossible in Asimov's universe to create a robot that was without them.

In real life, we know how trivial it would be to completely overwrite the operating system with arbritary software.

Just because it's a robot doesn't mean it's a Strong AI that would be capable of understanding the Three Laws.

(Let alone that the Three Laws are necessarily the best core ethical code to stick in your bipedal sentient robots.)

Applications are open for YC Winter 2018

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact