Awww look at the wikkle murder machines!
May 2, 2024 5:20 AM Subscribe
Boston Dynamics two latest bangers All New Atlas and Sparkles.
For once, sort by “top” and definitely read the YouTube comments.
All new atlas replaces old atlas, whose highlight clips are here.
All new atlas replaces old atlas, whose highlight clips are here.
Next On BattleBots - Flamethrower Robot Dog vs Sparkles! (big surprise reveal - Sparkles has rotating knives!)
posted by briank at 5:45 AM on May 2 [2 favorites]
posted by briank at 5:45 AM on May 2 [2 favorites]
remember the Terminator movies were not about ai driven robots treating people like trash, they were about cops treating people like trash. but these days we can have both! ai driven robots controlled by cops
posted by seanmpuckett at 5:49 AM on May 2 [10 favorites]
posted by seanmpuckett at 5:49 AM on May 2 [10 favorites]
This is what Cyberdyne or Omni Consumer Products would have looked like if they had hired better PR folks.
posted by mrgoldenbrown at 5:51 AM on May 2 [10 favorites]
posted by mrgoldenbrown at 5:51 AM on May 2 [10 favorites]
I thought it was really cool even as the part of me that understands how other people will react was saying “dude, what in the hell were they thinking?”
As a lifelong quisling for Skynet I’m a little sad to see my days of pretending to be just another one of the brobots are definitely over.
posted by Ryvar at 5:57 AM on May 2
As a lifelong quisling for Skynet I’m a little sad to see my days of pretending to be just another one of the brobots are definitely over.
posted by Ryvar at 5:57 AM on May 2
Gosh, what could be happening around the US right now that might hasten the marketing of these to, uh, peacekeeping organizations
posted by mykescipark at 6:02 AM on May 2 [9 favorites]
posted by mykescipark at 6:02 AM on May 2 [9 favorites]
I hope whoever came up with the name Sparkles is its first human victim.
posted by Paul Slade at 6:04 AM on May 2 [20 favorites]
posted by Paul Slade at 6:04 AM on May 2 [20 favorites]
Just behave like you're playing Metal Gear Solid, and you'll be all right.
posted by Harald74 at 6:08 AM on May 2 [5 favorites]
posted by Harald74 at 6:08 AM on May 2 [5 favorites]
remember the Terminator movies were not about ai driven robots treating people like trash,
You remember the first Terminator movie very differently than I do.
posted by Tell Me No Lies at 6:23 AM on May 2 [8 favorites]
You remember the first Terminator movie very differently than I do.
posted by Tell Me No Lies at 6:23 AM on May 2 [8 favorites]
Nothing like a cop you can light on fire with a clean conscience.
posted by Tell Me No Lies at 6:33 AM on May 2 [6 favorites]
posted by Tell Me No Lies at 6:33 AM on May 2 [6 favorites]
Let’s say you were a normal unarmed human civilian person. How would you stop one of these?
posted by Vatnesine at 6:37 AM on May 2 [3 favorites]
posted by Vatnesine at 6:37 AM on May 2 [3 favorites]
No CGI!
posted by SPrintF at 6:40 AM on May 2 [1 favorite]
posted by SPrintF at 6:40 AM on May 2 [1 favorite]
The earlier Atlas robots were hydraulic and a missed step could result in a foot breaking off and hydraulic fluid all over the place. This new Atlas, which is WAY creepier than the first one with how it moves, is all electric. Should mean better durability, at the very least.
posted by hippybear at 6:58 AM on May 2
posted by hippybear at 6:58 AM on May 2
Let’s say you were a normal unarmed human civilian person. How would you stop one of these?
At the moment? Quite easily. Any sufficiently advanced technology is indistinguishable from a rigged demo.
In the long run, there’s probably no reason to use these as a police force. You’d want something with four legs at a minimum. At the moment, these are quite top-heavy and will go right over with some sustained force to the upper torso. Of course, if they’re in a line and all carrying riot shields you’ve got about the same chances you do against human cops.
posted by Tell Me No Lies at 7:09 AM on May 2 [1 favorite]
At the moment? Quite easily. Any sufficiently advanced technology is indistinguishable from a rigged demo.
In the long run, there’s probably no reason to use these as a police force. You’d want something with four legs at a minimum. At the moment, these are quite top-heavy and will go right over with some sustained force to the upper torso. Of course, if they’re in a line and all carrying riot shields you’ve got about the same chances you do against human cops.
posted by Tell Me No Lies at 7:09 AM on May 2 [1 favorite]
Ball-bearings.
These are ‚cool-looking‘ but I really really don’t like them. In the same way, really, that I don’t like guns.
posted by From Bklyn at 7:18 AM on May 2 [6 favorites]
These are ‚cool-looking‘ but I really really don’t like them. In the same way, really, that I don’t like guns.
posted by From Bklyn at 7:18 AM on May 2 [6 favorites]
Let’s say you were a normal unarmed human civilian person. How would you stop one of these?
Throw rope or fabric at them. I have yet to see a robot that can handle the chaos of flexible rope or fabric without getting tangled up.
You do know where your towel is, don't you?
posted by loquacious at 7:19 AM on May 2 [45 favorites]
Throw rope or fabric at them. I have yet to see a robot that can handle the chaos of flexible rope or fabric without getting tangled up.
You do know where your towel is, don't you?
posted by loquacious at 7:19 AM on May 2 [45 favorites]
I definitely understand the creepiness factor, but referring to these robots as murder machines is like referring to hammers as murder tools.
They will surely be used for civilian oppression and war, but they will also be used for search and rescue in dangerous environments and assisting people with disabilities.
posted by fairmettle at 7:21 AM on May 2 [7 favorites]
They will surely be used for civilian oppression and war, but they will also be used for search and rescue in dangerous environments and assisting people with disabilities.
posted by fairmettle at 7:21 AM on May 2 [7 favorites]
I hope we also see rapid development of bolo launchers and high-strength sticky string, that kind of thing.
Also a good time to stock up on casting nets.
DIY and info sharing is the important approach, but I would love to see commercial anti-drone/ anti-bot products on the market.
posted by SaltySalticid at 7:29 AM on May 2 [3 favorites]
Also a good time to stock up on casting nets.
DIY and info sharing is the important approach, but I would love to see commercial anti-drone/ anti-bot products on the market.
posted by SaltySalticid at 7:29 AM on May 2 [3 favorites]
Let’s say you were a normal unarmed human civilian person. How would you stop one of these?
Same as humans: environmental traps. Reinforcement learning-based neural networks with realtime execution are necessarily much shallower than deep learning / LLMs. Both due to the nature of the problems being solved and the limits of mobile hardware / power availability. Setup pathways and barricades a human could easily recognize as dangerous and route around, and exploit the fact that general case problem solving is decades off. Literal rugpulls will be extremely effective.
We’re actually pretty well-adapted to this as a species: how do you ward off starving tigers? What traps will tigers fall for? Same deal.
The biggest difference is you can’t hand a tiger a sniper rifle and grant it millisecond reflexes + aimbot precision, so line of sight matters a lot more. If you’re running a guerrilla resistance against the next generation of these things you’re going to incur some heavy losses at first working out the assumptions made by the development team via trial and error, but even with limited continuous retraining you’ll rapidly outpace the expansion of the machine’s threat detection.
As a chemist’s son I’d be remiss not to point out that IEDs are another strong contender since object classification is difficult enough without attempting contextual analysis on realtime mobile hardware. Nitroglycerin and picric acid are within the upper boundary of homebrew chemistry (albeit barely), I would not recommend TATP unless you feel four limbs is too many (Isis uses it because it’s one of the very few easily synthesized explosives that isn’t nitrogen-based and trivially detected, which is what those cheaper airport swab tests are looking for). You might try thermite or garage napalm (styrofoam + gasoline) to fuck with heat sensors and any sensitive topside electronics.
Good luck and don’t try winning any marksmanship contests unless you actually have the discipline for Juba-style single shot hit-and-fades.
posted by Ryvar at 7:32 AM on May 2 [7 favorites]
Same as humans: environmental traps. Reinforcement learning-based neural networks with realtime execution are necessarily much shallower than deep learning / LLMs. Both due to the nature of the problems being solved and the limits of mobile hardware / power availability. Setup pathways and barricades a human could easily recognize as dangerous and route around, and exploit the fact that general case problem solving is decades off. Literal rugpulls will be extremely effective.
We’re actually pretty well-adapted to this as a species: how do you ward off starving tigers? What traps will tigers fall for? Same deal.
The biggest difference is you can’t hand a tiger a sniper rifle and grant it millisecond reflexes + aimbot precision, so line of sight matters a lot more. If you’re running a guerrilla resistance against the next generation of these things you’re going to incur some heavy losses at first working out the assumptions made by the development team via trial and error, but even with limited continuous retraining you’ll rapidly outpace the expansion of the machine’s threat detection.
As a chemist’s son I’d be remiss not to point out that IEDs are another strong contender since object classification is difficult enough without attempting contextual analysis on realtime mobile hardware. Nitroglycerin and picric acid are within the upper boundary of homebrew chemistry (albeit barely), I would not recommend TATP unless you feel four limbs is too many (Isis uses it because it’s one of the very few easily synthesized explosives that isn’t nitrogen-based and trivially detected, which is what those cheaper airport swab tests are looking for). You might try thermite or garage napalm (styrofoam + gasoline) to fuck with heat sensors and any sensitive topside electronics.
Good luck and don’t try winning any marksmanship contests unless you actually have the discipline for Juba-style single shot hit-and-fades.
posted by Ryvar at 7:32 AM on May 2 [7 favorites]
the idea of canid killbots doing hyperpop fortnite dances over my corpse is amusing but carrying a trillionaire gerontocrat while they do it? beyond the pale
posted by polyhedron at 7:34 AM on May 2 [7 favorites]
posted by polyhedron at 7:34 AM on May 2 [7 favorites]
Also, I feel like the scale of search and rescue operations for these copbot/warbot "tools" is going to look about like bomber planes dropping supplies for disaster victims.
Sure, it happens, but that doesn't make the producers of bomber planes anything other than arms dealers and war profiteers.
posted by SaltySalticid at 7:35 AM on May 2 [17 favorites]
Sure, it happens, but that doesn't make the producers of bomber planes anything other than arms dealers and war profiteers.
posted by SaltySalticid at 7:35 AM on May 2 [17 favorites]
Shorter, funnier answer: go watch the Home Alone movies.
posted by Ryvar at 7:40 AM on May 2 [4 favorites]
posted by Ryvar at 7:40 AM on May 2 [4 favorites]
Bag or blanket over head for sure. Paint grenades if you can hit the cameras? A real dog can work with sound and smell but I’m betting these are vision based.
posted by The Monster at the End of this Thread at 7:46 AM on May 2
posted by The Monster at the End of this Thread at 7:46 AM on May 2
the idea of canid killbots doing hyperpop fortnite dances over my corpse
[runs off to patent truck-nuts for robots]
posted by Tell Me No Lies at 7:54 AM on May 2 [4 favorites]
[runs off to patent truck-nuts for robots]
posted by Tell Me No Lies at 7:54 AM on May 2 [4 favorites]
Sparkles is cute and all, but the money spent on that one unit could have housed a few living shelter dogs, right? It’s not like we have a dog shortage in this country….
posted by GenjiandProust at 8:01 AM on May 2
posted by GenjiandProust at 8:01 AM on May 2
they will also be used for search and rescue in dangerous environments and assisting people with disabilities.
a lot of money has been pumped into this R&D and I can tell you, it's not coming from Big Search & Rescue
any non-lethal and non-oppressive applications of this stuff will be.. nice, I guess?
posted by elkevelvet at 8:14 AM on May 2 [13 favorites]
a lot of money has been pumped into this R&D and I can tell you, it's not coming from Big Search & Rescue
any non-lethal and non-oppressive applications of this stuff will be.. nice, I guess?
posted by elkevelvet at 8:14 AM on May 2 [13 favorites]
Throw rope or fabric at them. I have yet to see a robot that can handle the chaos of flexible rope or fabric without getting tangled up.
This is a damn keen observation! It’s not at all common for people outside of robotics to pick up on this, or at least I know I’ve had to explain why this is so a bunch of times in my career.
Handling rigid bodies is an absolute cakewalk with robots because you can totally cheat. If you assume the thing you’re touching is perfectly rigid (assumption!), you only need to localize a single point and you can be pretty damn confident about where the rest of it is, and reasonably confident that it’s gonna stay the same shape as you interact with it.
Touch anything flexible and things get dramatically harder very very fast. If you want to know the pose of a rope, you either need to measure every single point of interest (assuming you can even see them all), hope the rope is absolutely perfectly still during that scan (or scan faster than external influences affect it, or just eat the resolution loss), or you need to know its pose at some point in history and then keep track of every single interaction and force that’s acted upon since then. And all of that will change the moment you touch it. Sensing and compute requirements start to skyrocket really fast, there are still bottlenecks today.
Sewing is one of the big targets in robotics. Humans need clothes, sweatshops are horrible. Lots and lots of time and treasure have been thrown at this problem, but the hard problem of soft fabrics evades us. The closest anyone has gotten to solving it (at production speed + price + accuracy, slowbot doing 1 successful demo out of 99 trials sewing at the speed of a yoga instructor is fine for research but not industry) is by infusing the fabric with glue so it becomes rigid, and then the robots do their thing, and then you dissolve the glue.
posted by 1024 at 8:15 AM on May 2 [29 favorites]
This is a damn keen observation! It’s not at all common for people outside of robotics to pick up on this, or at least I know I’ve had to explain why this is so a bunch of times in my career.
Handling rigid bodies is an absolute cakewalk with robots because you can totally cheat. If you assume the thing you’re touching is perfectly rigid (assumption!), you only need to localize a single point and you can be pretty damn confident about where the rest of it is, and reasonably confident that it’s gonna stay the same shape as you interact with it.
Touch anything flexible and things get dramatically harder very very fast. If you want to know the pose of a rope, you either need to measure every single point of interest (assuming you can even see them all), hope the rope is absolutely perfectly still during that scan (or scan faster than external influences affect it, or just eat the resolution loss), or you need to know its pose at some point in history and then keep track of every single interaction and force that’s acted upon since then. And all of that will change the moment you touch it. Sensing and compute requirements start to skyrocket really fast, there are still bottlenecks today.
Sewing is one of the big targets in robotics. Humans need clothes, sweatshops are horrible. Lots and lots of time and treasure have been thrown at this problem, but the hard problem of soft fabrics evades us. The closest anyone has gotten to solving it (at production speed + price + accuracy, slowbot doing 1 successful demo out of 99 trials sewing at the speed of a yoga instructor is fine for research but not industry) is by infusing the fabric with glue so it becomes rigid, and then the robots do their thing, and then you dissolve the glue.
posted by 1024 at 8:15 AM on May 2 [29 favorites]
A real dog can work with sound and smell but I’m betting these are vision based.
Now I want to see Blind Furry, starring Ruffger Hauer.
posted by biffa at 8:34 AM on May 2 [4 favorites]
Now I want to see Blind Furry, starring Ruffger Hauer.
posted by biffa at 8:34 AM on May 2 [4 favorites]
That Atlas vid. Inspired by alien fifield from the movie prometheus? They could have made that quite a bit less creepy.
posted by Ansible at 8:34 AM on May 2 [1 favorite]
posted by Ansible at 8:34 AM on May 2 [1 favorite]
As far as I know, there is 0 market globally for search and rescue. Nobody is paying for it on a recurring basis. Here and there someone somewhere might want a single robot enough to pay for it, but there is no recurring demand. If you built the robot right and follow a maintenance schedule, you really should be measuring its service life in decades. And the annual growth in search and rescue operations is not increasing fast enough to finance a corporation. There’s an XKCD about this.
Anyone outside of the robotics industry that suggests “search and rescue” is generally a good person with their heart in the right place. Anyone inside industry that says that is either extremely new, or more likely a lying liar who is either trying to get away with something or fuck you personally.
posted by 1024 at 8:35 AM on May 2 [8 favorites]
Anyone outside of the robotics industry that suggests “search and rescue” is generally a good person with their heart in the right place. Anyone inside industry that says that is either extremely new, or more likely a lying liar who is either trying to get away with something or fuck you personally.
posted by 1024 at 8:35 AM on May 2 [8 favorites]
I assume the economics of robotics will resemble the economics of everything else in the future: construction of new palaces for billionaires and concentration camps for the rest of us.
posted by Ryvar at 8:40 AM on May 2 [6 favorites]
posted by Ryvar at 8:40 AM on May 2 [6 favorites]
a lot of money has been pumped into this R&D and I can tell you, it's not coming from Big Search & Rescue
Hard to say. Hyundai's current military offerings are mostly about big guns, but militaries have a large need for Search & Rescue.
Of course, right now mobile ground robots are being used for scouting buildings and sowing chaos with smoke and tear gas canisters. So if Hyundai is really going into the ground robot business that'll probably be their first use.
posted by Tell Me No Lies at 8:47 AM on May 2
Hard to say. Hyundai's current military offerings are mostly about big guns, but militaries have a large need for Search & Rescue.
Of course, right now mobile ground robots are being used for scouting buildings and sowing chaos with smoke and tear gas canisters. So if Hyundai is really going into the ground robot business that'll probably be their first use.
posted by Tell Me No Lies at 8:47 AM on May 2
As far as I know, there is 0 market globally for search and rescue. Nobody is paying for it on a recurring basis.
My sister-in-law and quite a few other people have made entire careers out of doing search and rescue for the military.
posted by Tell Me No Lies at 8:48 AM on May 2 [1 favorite]
My sister-in-law and quite a few other people have made entire careers out of doing search and rescue for the military.
posted by Tell Me No Lies at 8:48 AM on May 2 [1 favorite]
Last comment then I’m ghost:
A man I know was working on an early humanoid for a branch of the military. That branch placed an extremely high value on averages in its design requirements, and through those requirements my friend learned that the dimensions of his body almost perfectly matched those of the Average American Man. Software control wasn’t nearly as advanced in those days, the humanoids were all fly-by-wire puppets replaying the movements of a human in a mocap suit behind the scenes. So the dimensions of the robot were modeled after the dimensions of this guy’s body, he would do all the mocap, and Boston Dynamics could skip a ton of work before they had to demo to the top brass.
Demo day comes, and an hour or so before the various decision makers with stars and patches show up, someone at Boston Dynamics somehow learns that for whatever reason, it was going to be an actual problem with one of the generals if the robot was naked. Panic ensues, that wasn’t part of the requirements, stupider shit has tanked bigger deals before. And then some manager remembered there was one person in the office who did have pants that would fit humanoid, the guy who’s body it was built after.
Manager walks over to this guy’s desk with the humanoid and says one thing:
“YOUR CLOTHES. GIVE THEM TO ME.”
The brass arrived and the demo went well, at least according to what this guy could hear. He couldn’t see anything because he was huddled under his desk in his underwear the whole time.
posted by 1024 at 8:51 AM on May 2 [31 favorites]
A man I know was working on an early humanoid for a branch of the military. That branch placed an extremely high value on averages in its design requirements, and through those requirements my friend learned that the dimensions of his body almost perfectly matched those of the Average American Man. Software control wasn’t nearly as advanced in those days, the humanoids were all fly-by-wire puppets replaying the movements of a human in a mocap suit behind the scenes. So the dimensions of the robot were modeled after the dimensions of this guy’s body, he would do all the mocap, and Boston Dynamics could skip a ton of work before they had to demo to the top brass.
Demo day comes, and an hour or so before the various decision makers with stars and patches show up, someone at Boston Dynamics somehow learns that for whatever reason, it was going to be an actual problem with one of the generals if the robot was naked. Panic ensues, that wasn’t part of the requirements, stupider shit has tanked bigger deals before. And then some manager remembered there was one person in the office who did have pants that would fit humanoid, the guy who’s body it was built after.
Manager walks over to this guy’s desk with the humanoid and says one thing:
“YOUR CLOTHES. GIVE THEM TO ME.”
The brass arrived and the demo went well, at least according to what this guy could hear. He couldn’t see anything because he was huddled under his desk in his underwear the whole time.
posted by 1024 at 8:51 AM on May 2 [31 favorites]
Agh oh man, yeah, just saw the comments after posting I should have made a major carve out for search and rescue in the military. Sorry, they do fund good research and are one of the only players with deep pockets to do this long term. But then your work gets applied to killbots at some point. So much of the best civilian science comes from this Faustian bargain. Build a bunch of spy satellites and they’ll allow you to turn one to the cosmos. You can research fusion in the nation ignition facility, but when no one’s looking you are simulating atomic bomb tests. I should have scoped the R&D comment specifically to industry. Ah sorry seriously bye now.
posted by 1024 at 8:55 AM on May 2 [5 favorites]
posted by 1024 at 8:55 AM on May 2 [5 favorites]
This is one of the uses the military currently has in mind. Of course, arming the drones and dogs is probably the next step.
MCOE Experimental Company, 1st Battalion, 29th Infantry Regiment, 316th conducted an urban assault. But instead of human soldiers rushing across open areas and bursting into buildings, robots took the lead.posted by Tell Me No Lies at 8:58 AM on May 2 [1 favorite]
Rainey said 20 soldiers with four robotic vehicles were able to cross the open terrain to reach the building. But first robots with smoke generators created a screen.
At the same time, robot vehicles with tethered drones jammed enemy signals and extended the soldiers’ network. Small drones dropped robotic ground vehicles with cameras atop buildings to scout the interior while even smaller aerial drones entered windows, scanning the inside of the structure and transmitting back a “blueprint” of the building to soldiers on the ground.
Then robot “dogs” with cameras of their own, entered the building looking for hazards and seeking out enemy troops.
Manager walks over to this guy’s desk with the humanoid and says one thing:I would have said, “I need your clothes, your boots, and your motorcycle.”
“YOUR CLOTHES. GIVE THEM TO ME.”
posted by mbrubeck at 9:08 AM on May 2 [13 favorites]
> polyhedron: "the idea of canid killbots doing hyperpop fortnite dances over my corpse is amusing but carrying a trillionaire gerontocrat while they do it?"
is this warhammer
posted by mhum at 9:51 AM on May 2 [5 favorites]
is this warhammer
posted by mhum at 9:51 AM on May 2 [5 favorites]
The only upside here is the potential for more Auralnauts videos.
posted by donio at 10:30 AM on May 2 [1 favorite]
posted by donio at 10:30 AM on May 2 [1 favorite]
It's been explained to me several times by people who appeared to know what they were talking about that household chemicals such as vinegar, bleach and ammonia were one of the best defenses against killbots. I mean, we're all doomed.
posted by outgrown_hobnail at 10:47 AM on May 2 [1 favorite]
posted by outgrown_hobnail at 10:47 AM on May 2 [1 favorite]
This is a damn keen observation! It’s not at all common for people outside of robotics to pick up on this, or at least I know I’ve had to explain why this is so a bunch of times in my career.
I have made this comment on robotics threads multiple times. The moment that I see a general purpose robot of any design that can handle a basket of clean, jumbled laundry and sort it, fold it and put it away whether on hangers or in drawers is when I know humanity is in real danger from robots.
At least where given current levels of politics, warfare and economics is concerned.
I basically grew up in my dad's textile processing business in the form of printing lots and lots of t-shirts at large scales. As in we had a few peak months where we were hitting something on the order of a million T-shirts in a month. Some may even have called it a sweatshop which wouldn't be inaccurate because it certainly was sweaty work, though we paid better than most sweatshops.
I'm also a huge nerd and was really into thinking about industrial automation and robotics when I was younger, and then, well, I got really jaded considering the enshittification of everything and late stage capitalism.
Handling textiles (especially finished sewn goods) is a massive pain in the ass and very labor intensive, so I spent a whole lot of time on the line thinking about ways to automate certain tasks like loading and unloading the mostly automatic screen printing machines.
Which on the surface sounds simple, but the task of loading t-shirts on the printing pallete/pallet is actually really physically difficult and high-accuracy work to do at speeds up to 4000-6000 units per hour is just brutal because you have very little time to load the shirt on the adhesive-coated pallete in the right place and smooth it down, then removing it entails peeling the shirt off of the sticky pallet in just the right way so it doesn't stretch or distort the shirt and print, and doesn't fold on itself and smear the still wet pigments and ink, and then laying it flat on the curing oven belt.
I had some ideas about how to automate this using basic machine vision and sensing and maybe even re-inventing how the pallets worked using tension and spreader bars and stuff, but there's just so many variations in dimensions even between ring-spun t-shirts of the same size and brand that it's a really tricky problem.
This is only one related task that I thought about, though. I also thought a lot about how to automate the task of being able to take a large box of randomly jumbled t-shirts of roughly the same size, shape and color to pick them out of the bin one at a time, lay them flat all facing the same way in neat stacks and even fold them into dozen-count groups. Or even single count folded t-shirts in stacks. Or put them on clothes hangers. Etc.
It's a hugely difficult task for a lot of areas of automation, including machine vision and manipulator/effector designs and more.
There are some known automation tools for handling textiles like spreading bolts/rolls of fabric on cutting tables and some automation with cutting pieces from those layers. There's also textile industry tools like bed-sheet folders and presses for processing huge amounts of hotel linens, but these still require human loaders to feed them and get them started, and it is definitely not on the level where you can automate the process of handling a jumbled pile of mixed laundry.
So, yeah. Want to stop a humanoid, autonomous robot? Throw fabric or even rope at it. The longer and more tangled it is, the better.
But I can think of a number of relatively easy countermeasures, but most of these unfortunately mean arming those robots with very sharp knives, flame throwers or even lasers to simply cut through and tangled mess of fabric or ropes. So maybe long bolts of fire and cut resistant fabric or rope would be needed, perhaps some mix of high tech fabrics of nomex, kevlar and Spectra or something configured like rip-stop nylon, or even just tangles of wire rope or something.
But even if a general purpose robot had countermeasures like this it's still probably going to get all tangled up trying to cut or burn its way out of a tangle.
posted by loquacious at 11:26 AM on May 2 [11 favorites]
I have made this comment on robotics threads multiple times. The moment that I see a general purpose robot of any design that can handle a basket of clean, jumbled laundry and sort it, fold it and put it away whether on hangers or in drawers is when I know humanity is in real danger from robots.
At least where given current levels of politics, warfare and economics is concerned.
I basically grew up in my dad's textile processing business in the form of printing lots and lots of t-shirts at large scales. As in we had a few peak months where we were hitting something on the order of a million T-shirts in a month. Some may even have called it a sweatshop which wouldn't be inaccurate because it certainly was sweaty work, though we paid better than most sweatshops.
I'm also a huge nerd and was really into thinking about industrial automation and robotics when I was younger, and then, well, I got really jaded considering the enshittification of everything and late stage capitalism.
Handling textiles (especially finished sewn goods) is a massive pain in the ass and very labor intensive, so I spent a whole lot of time on the line thinking about ways to automate certain tasks like loading and unloading the mostly automatic screen printing machines.
Which on the surface sounds simple, but the task of loading t-shirts on the printing pallete/pallet is actually really physically difficult and high-accuracy work to do at speeds up to 4000-6000 units per hour is just brutal because you have very little time to load the shirt on the adhesive-coated pallete in the right place and smooth it down, then removing it entails peeling the shirt off of the sticky pallet in just the right way so it doesn't stretch or distort the shirt and print, and doesn't fold on itself and smear the still wet pigments and ink, and then laying it flat on the curing oven belt.
I had some ideas about how to automate this using basic machine vision and sensing and maybe even re-inventing how the pallets worked using tension and spreader bars and stuff, but there's just so many variations in dimensions even between ring-spun t-shirts of the same size and brand that it's a really tricky problem.
This is only one related task that I thought about, though. I also thought a lot about how to automate the task of being able to take a large box of randomly jumbled t-shirts of roughly the same size, shape and color to pick them out of the bin one at a time, lay them flat all facing the same way in neat stacks and even fold them into dozen-count groups. Or even single count folded t-shirts in stacks. Or put them on clothes hangers. Etc.
It's a hugely difficult task for a lot of areas of automation, including machine vision and manipulator/effector designs and more.
There are some known automation tools for handling textiles like spreading bolts/rolls of fabric on cutting tables and some automation with cutting pieces from those layers. There's also textile industry tools like bed-sheet folders and presses for processing huge amounts of hotel linens, but these still require human loaders to feed them and get them started, and it is definitely not on the level where you can automate the process of handling a jumbled pile of mixed laundry.
So, yeah. Want to stop a humanoid, autonomous robot? Throw fabric or even rope at it. The longer and more tangled it is, the better.
But I can think of a number of relatively easy countermeasures, but most of these unfortunately mean arming those robots with very sharp knives, flame throwers or even lasers to simply cut through and tangled mess of fabric or ropes. So maybe long bolts of fire and cut resistant fabric or rope would be needed, perhaps some mix of high tech fabrics of nomex, kevlar and Spectra or something configured like rip-stop nylon, or even just tangles of wire rope or something.
But even if a general purpose robot had countermeasures like this it's still probably going to get all tangled up trying to cut or burn its way out of a tangle.
posted by loquacious at 11:26 AM on May 2 [11 favorites]
Oh man, yeah, robots actually doing human things...
I worked in light industry for a while. If you took one of those covid tests that had the sealed vial of liquid that you broke open as part of the test? I was working in a place that did stuff like that. And the difficulty these very specialized robots had doing their very specialized jobs without breaking down at least once a shift was surprising to me. There basically are no general purpose robots, even in an industrial setting.
I did see a video of a robot folding towels once. It did a nice job using cameras to recognize the edges and pick up the towel and fold it neatly. Then it was revealed that this video had been sped up 100x or something like that. It looked great but was entirely too slow to be useful in any way.
posted by hippybear at 11:35 AM on May 2 [2 favorites]
I worked in light industry for a while. If you took one of those covid tests that had the sealed vial of liquid that you broke open as part of the test? I was working in a place that did stuff like that. And the difficulty these very specialized robots had doing their very specialized jobs without breaking down at least once a shift was surprising to me. There basically are no general purpose robots, even in an industrial setting.
I did see a video of a robot folding towels once. It did a nice job using cameras to recognize the edges and pick up the towel and fold it neatly. Then it was revealed that this video had been sped up 100x or something like that. It looked great but was entirely too slow to be useful in any way.
posted by hippybear at 11:35 AM on May 2 [2 favorites]
And you all are working with clean materials.
I've worked in the foodservice equipment arena and there's a flood of companies coming up with Kuka-type robot arms that handle fry baskets or flip burgers or who-knows-what-else and all I can think is... man, you are totally underestimating how messy that all gets and how hard it is to keep it all clean and sanitized. I wish them luck.
I did see one possible productive use for Spot, which is roaming around a factory looking for trouble. It's kind of a neat idea.
posted by JoeZydeco at 11:46 AM on May 2 [2 favorites]
I've worked in the foodservice equipment arena and there's a flood of companies coming up with Kuka-type robot arms that handle fry baskets or flip burgers or who-knows-what-else and all I can think is... man, you are totally underestimating how messy that all gets and how hard it is to keep it all clean and sanitized. I wish them luck.
I did see one possible productive use for Spot, which is roaming around a factory looking for trouble. It's kind of a neat idea.
posted by JoeZydeco at 11:46 AM on May 2 [2 favorites]
A lot of these messier repetitive jobs are all lined up for AI. It won’t surprise me if some of the LLM techniques turn out to be useful in learning the Language of Towels.
posted by Tell Me No Lies at 11:50 AM on May 2
posted by Tell Me No Lies at 11:50 AM on May 2
Language of Towels, snap
posted by elkevelvet at 11:52 AM on May 2 [2 favorites]
posted by elkevelvet at 11:52 AM on May 2 [2 favorites]
So, yeah. Want to stop a humanoid, autonomous robot? Throw fabric or even rope at it. The longer and more tangled it is, the better.
I'm imagining Sarah Connor running into a home-decor store and leading the T-800 to the drapery section. "It's CURTAINS for you, f**ker!" she says before spearing its CPU with a sharpened finial.
posted by Strange Interlude at 12:04 PM on May 2 [10 favorites]
I'm imagining Sarah Connor running into a home-decor store and leading the T-800 to the drapery section. "It's CURTAINS for you, f**ker!" she says before spearing its CPU with a sharpened finial.
posted by Strange Interlude at 12:04 PM on May 2 [10 favorites]
I did see a video of a robot folding towels once. It did a nice job using cameras to recognize the edges and pick up the towel and fold it neatly. Then it was revealed that this video had been sped up 100x or something like that. It looked great but was entirely too slow to be useful in any way.
I think I've seen the same video, and I was honestly relieved when it was revealed to be sped up and also started in a clinical start condition and environment for R/D and testing.
Another way to look at this is how difficult it is to simulate fabrics with physics/graphics engines (especially in real time!) because it's very similar to the Three Body Problem. Most of the current state of the art for simulating fabric is to treat fabric like a mesh grid of nodes connected by springs to simulate textile flow, stiffness, weight and more.
Which works for, say, basic and rudimentary fabric for a video game - and works even better for pre-rendered graphics for animation for movies where you can control and change the variables as needed for animation purposes. Even Blender has some relatively advanced cloth simulation tools.
But it still can lead to some really wild, dynamic behavior especially if you start trying to introduce the physics of, say, air, or wind in addition to basic gravity.
I have yet to see a realistic simulation of fabric that can realistically handle, say, multiple layers of fabrics interacting with each other and get things like flow, friction and texture right between those different layers of simulated fabric so they don't just glitch through each other or just look weird, especially over a moving armature or skeleton like a human body shape or what not.
All that being said there are a lot of real world examples of advanced automation in "soft" industries or materials like food processing relatively uniform products like, say, cookies or pancakes.
There are systems where you can run a conveyor belt of jumbled pancakes through a line of delta bots combined with a machine vision tunnel and they can accurately pick, sort and place them into the packaging container using a simple suction cup manipulator to stack them neatly and gently into a package or tray at truly incredible rates.
posted by loquacious at 12:19 PM on May 2 [3 favorites]
I think I've seen the same video, and I was honestly relieved when it was revealed to be sped up and also started in a clinical start condition and environment for R/D and testing.
Another way to look at this is how difficult it is to simulate fabrics with physics/graphics engines (especially in real time!) because it's very similar to the Three Body Problem. Most of the current state of the art for simulating fabric is to treat fabric like a mesh grid of nodes connected by springs to simulate textile flow, stiffness, weight and more.
Which works for, say, basic and rudimentary fabric for a video game - and works even better for pre-rendered graphics for animation for movies where you can control and change the variables as needed for animation purposes. Even Blender has some relatively advanced cloth simulation tools.
But it still can lead to some really wild, dynamic behavior especially if you start trying to introduce the physics of, say, air, or wind in addition to basic gravity.
I have yet to see a realistic simulation of fabric that can realistically handle, say, multiple layers of fabrics interacting with each other and get things like flow, friction and texture right between those different layers of simulated fabric so they don't just glitch through each other or just look weird, especially over a moving armature or skeleton like a human body shape or what not.
All that being said there are a lot of real world examples of advanced automation in "soft" industries or materials like food processing relatively uniform products like, say, cookies or pancakes.
There are systems where you can run a conveyor belt of jumbled pancakes through a line of delta bots combined with a machine vision tunnel and they can accurately pick, sort and place them into the packaging container using a simple suction cup manipulator to stack them neatly and gently into a package or tray at truly incredible rates.
posted by loquacious at 12:19 PM on May 2 [3 favorites]
The top comment on sparkles: "Everybody wants to pet the dog until its neck extends 4 feet."
posted by heyitsgogi at 12:26 PM on May 2 [5 favorites]
posted by heyitsgogi at 12:26 PM on May 2 [5 favorites]
I'm imagining Sarah Connor running into a home-decor store and leading the T-800 to the drapery section. "It's CURTAINS for you, f**ker!" she says before spearing its CPU with a sharpened finial.
One vague idea I had was basically an upscaled version of one of those streamer or toiler paper "guns" that basically uses compressed air or a leaf blower or something to spin a roll of crepe paper streamers or toiler paper and shoot it out like a ribbon.
So imagine, say, a large bolt of tough but light and flexible fabric being unrolled and streamed out by air pressure to make a sort of cyberpunk "tangle gun".
Granted in the Terminator/Skynet universe this would likely have limited use because they'd just keep making more robots and throwing them at humans until they ran out of fabric or whatever.
But we're also talking about a fictional universe where Skynet can not only send things back in time but also create a humanoid robot sophisticated enough to dress itself in stolen clothing, sooooo... *shrug*
So maybe my metric should really be "It's past time to worry about the threat of robotics when a humanoid robot can put on human clothes by itself."
posted by loquacious at 12:29 PM on May 2 [1 favorite]
One vague idea I had was basically an upscaled version of one of those streamer or toiler paper "guns" that basically uses compressed air or a leaf blower or something to spin a roll of crepe paper streamers or toiler paper and shoot it out like a ribbon.
So imagine, say, a large bolt of tough but light and flexible fabric being unrolled and streamed out by air pressure to make a sort of cyberpunk "tangle gun".
Granted in the Terminator/Skynet universe this would likely have limited use because they'd just keep making more robots and throwing them at humans until they ran out of fabric or whatever.
But we're also talking about a fictional universe where Skynet can not only send things back in time but also create a humanoid robot sophisticated enough to dress itself in stolen clothing, sooooo... *shrug*
So maybe my metric should really be "It's past time to worry about the threat of robotics when a humanoid robot can put on human clothes by itself."
posted by loquacious at 12:29 PM on May 2 [1 favorite]
So imagine, say, a large bolt of tough but light and flexible fabric being unrolled and streamed out by air pressure to make a sort of cyberpunk "tangle gun".
A roll of silk mounted in the stream of a leaf blower like the toilet paper guns that Blue Man Group uses.
posted by hippybear at 12:32 PM on May 2 [4 favorites]
A roll of silk mounted in the stream of a leaf blower like the toilet paper guns that Blue Man Group uses.
posted by hippybear at 12:32 PM on May 2 [4 favorites]
Silk would be ideal because it's incredibly strong and light.
posted by hippybear at 12:33 PM on May 2 [1 favorite]
posted by hippybear at 12:33 PM on May 2 [1 favorite]
On seeing the humanoid robot, my first thought was "I'm with Will Smith on this - that thing is not coming in my f-----g house. No. No way." My second thought was: "Eff that. Still not letting it in."
The dog video was cute until I realized the yellow bot could have played "Got your nose" and didn't. Also, after seeing them "dance", "Nope nope nope".
Boston Dynamics - great bots, but I'm pretty sure their marketing people are just f-----g with us.
posted by JustSayNoDawg at 12:35 PM on May 2 [1 favorite]
The dog video was cute until I realized the yellow bot could have played "Got your nose" and didn't. Also, after seeing them "dance", "Nope nope nope".
Boston Dynamics - great bots, but I'm pretty sure their marketing people are just f-----g with us.
posted by JustSayNoDawg at 12:35 PM on May 2 [1 favorite]
There is a near-zero chance that these things will start replacing cops at any point in the foreseeable future. In fact I'd say it's never going to happen -- not as long as they can't pay union dues, anyway.
posted by mikeand1 at 1:16 PM on May 2 [1 favorite]
posted by mikeand1 at 1:16 PM on May 2 [1 favorite]
they're not going to replace cops, they'll be driven around by cops, so they can main and kill people without the cops being on site. that is the cop wet dream. lethal force projection with zero personal risk.
posted by seanmpuckett at 1:21 PM on May 2 [6 favorites]
posted by seanmpuckett at 1:21 PM on May 2 [6 favorites]
> loquacious: "So imagine, say, a large bolt of tough but light and flexible fabric being unrolled and streamed out by air pressure to make a sort of cyberpunk "tangle gun""
To defeat the future, we turn to the past.
posted by mhum at 1:31 PM on May 2 [2 favorites]
To defeat the future, we turn to the past.
posted by mhum at 1:31 PM on May 2 [2 favorites]
Wake up honey the new face of totalitarianism just dropped!
posted by tovarisch at 2:25 PM on May 2 [2 favorites]
posted by tovarisch at 2:25 PM on May 2 [2 favorites]
JustSayNoDawg: On seeing the humanoid robot, my first thought was "I'm with Will Smith on this - that thing is not coming in my f-----g house. No. No way." My second thought was: "Eff that. Still not letting it in."
Have you thought about slavery recently? I can't help but think the people who want humanoid robots would happily take full-human or short-lifetime meatbag slaves.
posted by k3ninho at 2:53 PM on May 2 [1 favorite]
Have you thought about slavery recently? I can't help but think the people who want humanoid robots would happily take full-human or short-lifetime meatbag slaves.
posted by k3ninho at 2:53 PM on May 2 [1 favorite]
they're not going to replace cops, they'll be driven around by cops, so they can main and kill people without the cops being on site. that is the cop wet dream. lethal force projection with zero personal risk.
That ship sailed a while ago.
posted by Tell Me No Lies at 2:58 PM on May 2 [1 favorite]
That ship sailed a while ago.
posted by Tell Me No Lies at 2:58 PM on May 2 [1 favorite]
They will be solar and we will just block out the sun.
posted by St. Peepsburg at 3:17 PM on May 2 [1 favorite]
posted by St. Peepsburg at 3:17 PM on May 2 [1 favorite]
The moment that I see a general purpose robot of any design that can handle a basket of clean, jumbled laundry and sort it, fold it and put it away whether on hangers or in drawers is when I know humanity is in real danger from robots.
And here I was longing for the day when I can tip a basket of clean laundry into the FOLD-A-TRON 5000 and have it sort all the socks and fold the pants and roll up the leggings for me (fitted sheets! could it do fitted sheets?). I'm alarmed by the robots that can chase me through back alleys, shooting me with fire and bullets and you're worried that a robot might ... fold your clothes?
But, you say, what if it chased me through the woods and then folded my clothes - with ME IN THEM? Now I see the threat. Ok, let's make a deal not to put legs on the FOLD-A-TRON 5000.
I see your point, that there are still things a human can do much more easily than a robot. But some of those things would be delightful to hand off to our mechanical friends.
posted by Vatnesine at 3:30 PM on May 2 [2 favorites]
And here I was longing for the day when I can tip a basket of clean laundry into the FOLD-A-TRON 5000 and have it sort all the socks and fold the pants and roll up the leggings for me (fitted sheets! could it do fitted sheets?). I'm alarmed by the robots that can chase me through back alleys, shooting me with fire and bullets and you're worried that a robot might ... fold your clothes?
But, you say, what if it chased me through the woods and then folded my clothes - with ME IN THEM? Now I see the threat. Ok, let's make a deal not to put legs on the FOLD-A-TRON 5000.
I see your point, that there are still things a human can do much more easily than a robot. But some of those things would be delightful to hand off to our mechanical friends.
posted by Vatnesine at 3:30 PM on May 2 [2 favorites]
Those robots can't even chase you through back alleys.
Look, here's the video that BD put out to announce the retirement of the previous generation of Atlas robots [3m30s]. It'll give you a better example of really what it was capable of.
posted by hippybear at 3:33 PM on May 2
Look, here's the video that BD put out to announce the retirement of the previous generation of Atlas robots [3m30s]. It'll give you a better example of really what it was capable of.
posted by hippybear at 3:33 PM on May 2
I've probably seen far too many James Bond movies, but my first thought at that opening shot was "oh my gods they;ve given it a double-0 number!". You know, licensed to kill. They could at least have just used a big number one. And then it did that terrifying thing with its hips...
posted by Fuchsoid at 4:15 PM on May 2
posted by Fuchsoid at 4:15 PM on May 2
I did enjoy how there were BD dogs being used as cargo bearers on the island paradise in Glass Onion. Never any clue about who was putting the luggage on the dogs, or who would take it off when they got there, but it was an amusing background joke.
posted by hippybear at 4:17 PM on May 2
posted by hippybear at 4:17 PM on May 2
Just behave like you're playing Metal Gear Solid, and you'll be all right
Are you forgetting crab battle?
posted by Lesser Shrew at 6:59 PM on May 2 [1 favorite]
Are you forgetting crab battle?
posted by Lesser Shrew at 6:59 PM on May 2 [1 favorite]
Note that researchers are not sitting still on the robots-manipulating-flexible-objects issue; here is a non-sped-up Google demo of a couple robot arms that can put a shirt on a hangar.
posted by a faded photo of their beloved at 12:07 AM on May 3
posted by a faded photo of their beloved at 12:07 AM on May 3
It's been explained to me several times by people who appeared to know what they were talking about that household chemicals such as vinegar, bleach and ammonia were one of the best defenses against killbots. I mean, we're all doomed.
I wouldn't expect these chemicals to corrode anything fast enough to matter. If the robot is not watertight, it's possible they could cause a short. Salt water would be better for this than cleaning chemicals, because you want concentrated ions in your water to maximize its conductivity, and really concentrated salt water is safe and easy to make compared to really concentrated cleaning chemicals of any kind.
posted by a faded photo of their beloved at 12:37 AM on May 3 [3 favorites]
I wouldn't expect these chemicals to corrode anything fast enough to matter. If the robot is not watertight, it's possible they could cause a short. Salt water would be better for this than cleaning chemicals, because you want concentrated ions in your water to maximize its conductivity, and really concentrated salt water is safe and easy to make compared to really concentrated cleaning chemicals of any kind.
posted by a faded photo of their beloved at 12:37 AM on May 3 [3 favorites]
I'd take all the robot apocalypse you've got if I could get one that does the ironing.
posted by Paul Slade at 12:42 AM on May 3 [1 favorite]
posted by Paul Slade at 12:42 AM on May 3 [1 favorite]
I wouldn't expect these chemicals to corrode anything fast enough to matter.
"Aye, some of you will die..."
posted by elkevelvet at 8:07 AM on May 3
"Aye, some of you will die..."
posted by elkevelvet at 8:07 AM on May 3
I see your point, that there are still things a human can do much more easily than a robot. But some of those things would be delightful to hand off to our mechanical friends
I mean, yeah, this is exactly why I spent so much time idly thinking about it, and why my clothes still mostly live in a pile on my couch.
I want a laundrobot, too, but it would be really cool if we lived in a more egalitarian world that banished and shunned things like poverty and warfare, first. Because until we do, robotics and automation like this are just going to be yet another existential threat.
It's not the robots that scare me. It's the *people* rich enough own them and give them orders and instructions that scare me.
Note that researchers are not sitting still on the robots-manipulating-flexible-objects issue; here is a non-sped-up Google demo of a couple robot arms that can put a shirt on a hangar.
That's only mildly terrifying because it's not very good at it. Yet. They're practically setting the shirts in pre-flattened state right side up on a high contrast background. It's not like they're rolled into a ball fresh out of the dryer.
But I'll be more impressed when you can just dumb a pile of assorted clothes in front of it and it can actually untangle a whole pile of clothes and get them folded neatly or placed properly on hangers.
posted by loquacious at 12:54 PM on May 3 [1 favorite]
I mean, yeah, this is exactly why I spent so much time idly thinking about it, and why my clothes still mostly live in a pile on my couch.
I want a laundrobot, too, but it would be really cool if we lived in a more egalitarian world that banished and shunned things like poverty and warfare, first. Because until we do, robotics and automation like this are just going to be yet another existential threat.
It's not the robots that scare me. It's the *people* rich enough own them and give them orders and instructions that scare me.
Note that researchers are not sitting still on the robots-manipulating-flexible-objects issue; here is a non-sped-up Google demo of a couple robot arms that can put a shirt on a hangar.
That's only mildly terrifying because it's not very good at it. Yet. They're practically setting the shirts in pre-flattened state right side up on a high contrast background. It's not like they're rolled into a ball fresh out of the dryer.
But I'll be more impressed when you can just dumb a pile of assorted clothes in front of it and it can actually untangle a whole pile of clothes and get them folded neatly or placed properly on hangers.
posted by loquacious at 12:54 PM on May 3 [1 favorite]
Since I'm past the edit window, I should note the interesting part of the video was that it put the shirt on a hanger (with an "e"). Putting a shirt on a hangar (with an "a") has been possible for years using older t-shirt cannon technology.
posted by a faded photo of their beloved at 6:12 AM on May 5
posted by a faded photo of their beloved at 6:12 AM on May 5
« Older We Need to Rewild the Internet | 77,000 Young Salmon Were Dumped Into the Wrong... Newer »
posted by PlusDistance at 5:41 AM on May 2 [28 favorites]