/tech/ - Technology★

Welp, robots can finally move naturally now instead of being clunky as fuck like they've historically have been.

I'm not going to lie, I thought it would've been simply impossible from a practical engineering standpoint for robots THAT athletic to br built. But those absolute madmen did it. Now all it needs is for someone to install a moderm chatbot/"intelligent agent" software on it and it'll be just like sci-fi

>robots before mecha or even exoskeletons outside of experimentation

I just want the world to be more like patlabor

You do know they are making these to kill people?

Old news. (((They've))) had the tech to do this for decades now. Tell me when (((they))) start using death robots or something.

>>822656

This is why you carry a mini-emp or a taser with sharp as fuck prongs.

google owns them right?

>>822661

I thought they sold them back off

>>822661

They sold Boston Dynamics to a Japanese company SoftBank.

>>822654

Plans maybe, but unless you have evidence to back up your claim, I'll just go ahead and dismiss you.

What's the point of creating robots human-like?

Aren't there better choices? Humans don't seem like best structurally shaped things.

>>822932

It's probably sub-optimal but most of our environments and tools are designed for human body anyway. E.g. Spot Mini or Cassie most likely won't be able to climb ladders, while Atlas will be.

>>822634

>Now all it needs is for someone to install a moderm chatbot/"intelligent agent" software on it and it'll be just like sci-fi

You know how this will end up. I, for one, welcome our anime-girl Hitler overlords.

>>822931

Exoskeletons can have mechanical limits to avoid articulations bending the wrong way, and mechs won't bend since you are supposed to pilot them from inside a cabin. Mechs are impractical for other reasons, such as being so fuckhueg balancing them may prove difficult, or because the only upside of making them locally manned would be signal jammer immunity.

Gundam-style mechs are a pipe dream anyway. We will be more likely to design Tachikoma-style mechs than impractical bipedal stuff.

>>822932

>What's the point of creating robots human-like?

Srs?

>Do you even /robowaifu/ ?

The obvious military advantages just scream out here anon.

>>822947

>>822942

Hate to break it to ye guys, but Gundam mechs are more plausible than the AI that can think in human terms.

>>822546 (OP)

This has to be elaborate CGI.

>>822949

Both will literally happen before you die anon. Enjoy the ride.

>>822955

As a TD in the character animation field I can tell you it plainly has MOCAP written all over it. As a student in robotics and AI, I can guess it has a kinematic model built on a large set of takes by a single man. It's real anon.

>>822960

Given the balancing dynos, he was probably made to wear a backpack to simulate the weight of the robot's batteries, scaled ofc to the correct amounts. I'm betting an asian guy, probably Chinese.

>>822960

Motion Capturing definetly makes sense. What I wonder is how they generate that much power output (in the literal sense). If they are using electromagnetical actuators (what I suppose), then they need some big ass capacitors for supplying the jumping movements. Propably that thing is overheated after this video shot too and needs to cool down some time.

In general, I really admire what Boston Dynamics offers, technological-wise. But the overwhelming feelig I get is, that in some years, when the NWO is in place, these fuckers will hunt me down the streets because I accidentally whispered some wrongthink out loudly, aksing me to provide biometrical identification. If I don't comply, they arrest me and take me to cell 101.

>>822973

You've just outlined all the engineering requirements and tradeoffs in a single post anon. Well played.

As for being hunted down like dogs in the streets, tbqh I think, well, dogs would do a better job at it. Link related.

>>822656

>>822955

>>822960

That was my first thought, too. It moves like a human, not like a learning machine that has learned the most efficient ways to move by trail and error.

Then again, Boston Dynamics is an investment scam. Their main business is creating PR material to solicit money, not building robots. This could be a kinetic sculpture, designed to replicate movements from a mocap library. They have done this before with their running machines. In this case it can only perform the exact movements shown in the video, maybe some extra flourishes to show off to potential investors in person, and nothing more.

I wish there was more to it. But, they, or someone licensing their tech, would have started selling products a decade ago, if they were interested in creating products, or serious research, at all.

>>822975

You've got it anon.

>>822975

>This could be a kinetic sculpture, designed to replicate movements from a mocap library.

As a student of control theory I have to say, this is most likely impossible. If this is not CGI, it has to comply with the laws of physics. You can't even emagine how difficult it is to control (in the sense of closed feedback loops) simple motions, let alone landing a backflip without killing yourself (the robot). Not possible by just replicating the kinematics of a movement (if it is inherently unstable, like losing contact to the floor). I know they make extensive use of neural networks to accomplish difficult control tasks, but this one is a bit over the top.

>>822980

Only the training phase needs prohibitively costly computation anon. The runtime phase is relatively cheap and, even at that, easily off-loaded to external hardware in a dog-and-pony scenario such as this demo.

>>822982

Their tethered running machines are legit. I think you can even buy one of them. But they can only perform a small set of pre-programmed motions. Add some modern capacitors (they only have to last for a second or so) and modern processors and you are not too far of from what's shown in the video. It would still be useless, but it does look really neat in a youtube video.

>>822985

Actually, now that it's all under the control of the Nips, this shit will be made real. The national government has a yuge vested intredast in creating a robohumanoid service sector, particularly in the elderly care field.

>tl;dr

It may be fake now, but give it five years anon.

>>822988

>It may be fake now, but give it five years anon.

Sadly, that's the deal with those kind of companies. They always have the biggest parties, the best presentations and the friendliest representatives. But they always need just a few more years of research.

>>822988

>tldr-ing a one-liner

you need to go back

>The national government has a yuge vested intredast in creating a robohumanoid service sector

so far, so legit

>particularly in the elderly care field

Kek. Maybe the big pharma companies have an interest in keeping elderly people alive to farm off their pensions. The government only has one goal, which is total control. And these robots are a platform that seems it can just offer that.

>>822992

lolnou. the Internet, with all it's wild, zany educational opportunities, is a thing anon. try it someday.

>>822990

Fair points. But I'm quite certain SoftBank will take these guys down an entirely different path than DARPA was. Japan will make anime real, it's inevitable. The alternatives are simply too costly at this point for their nation's infrastructure and economy.

>>822980

>As a student of control theory

>>822960

>As a TD in the character animation field

>As a student in robotics and AI

As an X my like opinion somehow matters more. No, fuck off. You queers need to go back to Reddit and bugchase over there. I cannot stand the way you valley fags talk. We will judge you on what you say, not any credentials you bestow upon yourself like a desperate appeal to authority. Die in a fire.

>>822932

We were created in the image of our creator, it only makes sense that our creations are made in our image.

>>823086

Not really. I'd rather not repeat his mistakes. I'd go with some sort of tentacle monster, if I had to design a death machine meant to operate in human spaces.

>We

You meant "I" there, but you have a point. No input on the topic from you nonetheless.

>>822667

>>822659

>Nothing a 12ga slug can't fix.

Except an armor. The robot will be much much more precise and lethal for you at a 200m distance than you for him at 10m anyway.

>mini-emp

Will most likely include protection against that.

>>822974

Dogs get tired. Dogs don't see very well in the dark. Dogs can be killed more easily. Dogs can't have real time access to city CCTV. Dogs can't carry and fire guns. Dogs can't silently coordinate with each other in real time over large distances.

>It moves like a human, not like a learning machine that has learned the most efficient ways to move by trail and error.

Why would it? Humans have a relatively optimized walk already so it's far better to train this robot, which has a human-like anatomy, with human data.

>>823123

Brutal quadruped. If I saw that thing bearing down on me with a pair of miniguns at it's side I'd do a 360 and run tbh.

>>823082

Fair enough. Since I actually am a TD, I expect my understanding of both kinematics, system rigging and structural constraints for animatable systems somewhat outstrips the norm. Not to mention a wide swath of several other related topics to robotics in general. Don't like it? Prove me wrong on some point related point. Otherwise, kindly fuck off you'reself friend.

>>823111

>if I had to design a death machine meant to operate in human spaces.

>Not just raining down flaming rocks on the heads of evil men.

Plebeian.

>>823616

>which has a human-like anatomy

??

It would be best to let it learn by mistake in a simulated environment.

>>822651

I'll bet a shotgun blast close range would wreck that thing. There's also the issue of energy consumption. It has a long way to go before it resembles Arnold.

>>822947

>my family lives there

lost

Daily reminder that a robot is useless when it's proprietary. Only good robot is FLOSS robot.

>>823929

Absolutely. My first designs will all use only linux or other open sauce OS, the code will all be open sauce (probably BSD license) and all the printable parts and other designs will be freely available and open sauce too.

>>823929

I don't think open sourcing kill routines is a good idea unless you want Skynet to come to life sooner.

>>823934

>Not wanting to maximize the chances for someone to find a 0 day in Skynet and shut it down

As >>824170 says, if Skynet had been FLOSS, Terminator 3 would have never happened. Which, paradoxically, would have also eliminated Terminator 1 and 2.

Instead of that, Skynet was made by the NSA and the resistance got its ass rekt.