Could we sooner see overseas outsourcing of home chores? It'd be so much cheaper to have a human robotically fold my laundry remotely than to sponsor them with a visa. They can earn foreign...
It’s also notable that Neo’s actions are done with a combination of automation and teleoperation (where a human remotely controls Neo), so its actual autonomous capabilities are even less than what we see. [...] We noted above that many of 1X’s actions are in practice teleoperated, and most of the impressive-looking demonstrations of Tesla’s Optimus (such as shirt-folding) are also teleoperated.
Could we sooner see overseas outsourcing of home chores? It'd be so much cheaper to have a human robotically fold my laundry remotely than to sponsor them with a visa. They can earn foreign currency without leaving their loved ones. A sort of Da Vinci surgical system used to scramble eggs and put groceries away. Use the basic robot to get around the house, have the remote operator take over the hands to pass the dexterity check.
Here is a list of 21 dexterously demanding tasks that are relatively straightforward for a human to do, but I think would be extremely difficult for a robot to accomplish. [...] As with AI evals, ideally these would be done without the system being trained on these tasks specifically
Assemble a mechanical watch
Dude. I would say most people would be incapable of assembling or replacing the escapement assembly even after watching a few demos, without downright wrecking the watch or at the very least they might be able to cram it in but not balancing it properly. Without training or existing familiarity I would be very willing to bet almost no humans would be able to do this. Replace a quartz watch battery is a more fair task to ask of regular humans.
Roll a small lump of playdough into three smaller balls
We already have any number of similar machines that roll dough into precise smaller quantities. This one sounds easier if evals involve accessing a bigger database of what other robots know....right?
There is actually a pretty good scifi movie based on that premise called Sleep Dealer (Trailer): https://en.wikipedia.org/wiki/Sleep_Dealer
It'd be so much cheaper to have a human robotically fold my laundry remotely than to sponsor them with a visa. They can earn foreign currency without leaving their loved ones.
Sleep Dealer is a 2008 futuristic science fiction film directed by Alex Rivera.
Sleep Dealer depicts a dystopian future to explore ways in which technology both oppresses and connects migrants. A fortified wall has ended unauthorized Mexico-US immigration, but migrant workers are replaced by robots, remotely controlled by the same class of would-be emigrants.
I really wonder how a system like this would impact the US politically. Lots of blue collar manufacturing jobs have already been exported to countries where labor is cheaper since the cost to ship...
I really wonder how a system like this would impact the US politically. Lots of blue collar manufacturing jobs have already been exported to countries where labor is cheaper since the cost to ship the goods back is low. This has caused a ton of discontent with US citizens who would otherwise work in domestic manufacturing.
Construction and the service industry is now the best option for many blue collar workers in the US who want to make a decent living without a college education. If we use remotely controlled robots to fill those positions for less money, what is left for US workers?
I know UBI is the logical conclusion, but that will likely come after mass discontent/unrest because I don’t count on US legislators to be proactive about anything. Ideally, blue collar workers would just upskill and get a higher education, but not everyone wants to or is able to get a college degree.
I personally think that if you want to do well in society and society changes, you should try to adapt to the new reality. I also understand that when a small group of people refuses to or is unable to adapt to a changing economy, it’s generally their personal problem to deal with. When a significant percentage of the country struggles to deal with changes in the economy it becomes the government’s problem.
I am with you. As much as I love watching those Boston Dynamics videos of robots jumping around in construction sights, where is my cleaning robot or massage robot?!?! This is going to be like the...
I am with you. As much as I love watching those Boston Dynamics videos of robots jumping around in construction sights, where is my cleaning robot or massage robot?!?! This is going to be like the flying cars hopes and dreams of this generation. I want Rosey from the Jetsons, damn it.
I guess for me I learned a while ago that machines and automation, are not always for the machine to do something "better" or even at the same level of a person. Sometimes it is a machine that works good enough, to let the person do other things. Sometimes its to reduce the hazards associated with a human doing it (bomb diffuser, or repetitive tasks). I am mostly speaking about robots I used in labs before, they were not faster than a human doing the thing, but they allowed the humans to go do other things that only the humans could do. Automated machines are also generally more consistently precise than humans, so they are a good use case there; additionally, robots don't need sleep allowing them to perform things at a scale that a human cannot, but they don't do the individual thing better, necessarily.
Think about roombas. For me the amount of effort to make sure my floor is spotless so that the Roomba can operate without my attention, is so much work, at that point I'd rather clean the floor myself. It's like people who developed the roomba never cleaned in their lives, or were just the final person who mops a completely cleared out floor..."wow cleaning is so easy".
There's a reason automation and robots have not taken over the world, yet, and it's because at the end of the day they are not affordable over time (who maintains the robots and how expensive are those people, how replaceable are those people) and/or they don't do the job well enough (cleaning). Additionally, many people will literally do work for free. There is nothing cheaper than work for free.
Nobody does because Roombas suck. The Roborock Saros Z70 is where it's at. It actually does the thing you're talking about. It has an arm and picks up stuff from the floor. It's not that they...
Think about roombas.
Nobody does because Roombas suck. The Roborock Saros Z70 is where it's at. It actually does the thing you're talking about. It has an arm and picks up stuff from the floor. It's not that they didn't think about it when designing a robot vacuum, it's just very hard.
Even one without an arm is still a net benefit. It allows me to not do the vacuuming and mopping, but it does force me to pick stuff up and put it away.
It ties into your larger point, robots should exist to reduce human effort and my robot vacuum definitely does.
I somewhat hate these demos, because they set the wrong expectations. What they're showing off IS fascinating as hell, but every single move has been preprogrammed to an absurd degree. It's not...
As much as I love watching those Boston Dynamics videos of robots jumping around in construction sights, where is my cleaning robot or massage robot?!?!
I somewhat hate these demos, because they set the wrong expectations. What they're showing off IS fascinating as hell, but every single move has been preprogrammed to an absurd degree. It's not "just run over there" or "do a dance". It's a ton of very precise code telling it how to move its arms/legs/etc to get those outputs.
It's still VASTLY less code than literally hard coding it (which still would've been impossible up until recently), but the idea that these things are anywhere close to generic tasks, let alone cheaply, is 100% in flying car territory.
Both in the lack of tech, and likely the lack of practicality. There's a reason you don't have a stationary clothes folding maching in your house, even though factories have had them for decades. It's big, loud, and expensive. Only worth it if you're folding THOUSANDS of shirts, not your 20-100.
Are you distinguishing flying cars from air taxis? There are several companies nearly ready to offer air taxi service to the public although I don't know their price point.
Are you distinguishing flying cars from air taxis?
There are several companies nearly ready to offer air taxi service to the public although I don't know their price point.
Yes, although it somewhat depends on your definitions of air taxi, given that's something that has existed for some time for planes and helicopters. If you mean the quad copter style things,...
Yes, although it somewhat depends on your definitions of air taxi, given that's something that has existed for some time for planes and helicopters.
If you mean the quad copter style things, that's basically just a helicopter with different steps (maybe arguably better given how hellishly complicated helicopters are), and I don't think comes close to qualifying as something you can drive anywhere or use for point to point (if anything more equivalent to a bus or trolley).
Even more so, i sincerely doubt any of them are going to actually start any sort of service. The whole problem with ANY sort of flying car is just how insanely dangerous it is. Imagine if every single fender bender was fatal, that's the reality you're looking at. So of course we've built a MASSIVE system to make sure that everything in the air is super far apart, logs its EXACT path, follows explicit rules, and is constantly monitored by machines and humans both on the ground and in the air, and people still die.
In fact the most dangerous flying device you can get in is a helicopter. They fly low enough to the ground and have awkward enough aerodynamics that should something go wrong you have an insanely small window to HOPEFULLY fix it. Nothing about the new "air taxi's" i've seen addresses ANY of this.
The only reason i'm not certain is we have "self driving cars" that ALSO shouldn't be allowed given their lack of data, testing, and quality, and I see these air taxi's as a similar level of "90% hype with no thought for the actual legalities and issues this might cause", but apparently that's ok now.
If you've got an article that's got something saying otherwise I'm all ears, but everything I've seen sure as hell reminds me of https://en.wikipedia.org/wiki/Moller_M400_Skycar. Every air taxi copter thing i'm aware of only has the "Experimental" approval from the FAA, which basically means "not even remotely safe or close to it". Should they get any sort of actual approval from the FAA i'll keep my eyes open, but no one is "nearly ready" if they don't already have that.
They're kind of worse in virtually every way actually, aside from the complexity thing. They're less safe, because they lack the ability to autorotate, so if you have a total power loss, not only...
maybe arguably better given how hellishly complicated helicopters are
They're kind of worse in virtually every way actually, aside from the complexity thing. They're less safe, because they lack the ability to autorotate, so if you have a total power loss, not only do you just drop like a rock, you have zero control authority over the aircraft, because motors are the only thing giving you control authority, and they require power. They're less efficient because they use small props instead of large rotors. They're louder, because the blades need to spin much quicker to generate the same amount of lift, and there are more of them, and then obviously they have extremely limited flight time due to being electrically powered. There's a reason multirotors haven't supplanted helicopters for human transport in general. Those reasons haven't really changed just because you're charging fares for shorter trips now.
Yeah that was my expectation. Digging into the one linked it's supposedly somehow quieter, but it also appears to be using double blades per "wing" and also doing a V22 osprey thing where in...
Yeah that was my expectation. Digging into the one linked it's supposedly somehow quieter, but it also appears to be using double blades per "wing" and also doing a V22 osprey thing where in flight they rotate and it flies "forward".
Granted sharing anything with the V22 also doesn't bode well.
It's likely because it uses ducted fans, which are quieter than props/rotors, but are also way less efficient at the low speeds that a hovering aircraft operates at, and have a number of other...
It's likely because it uses ducted fans, which are quieter than props/rotors, but are also way less efficient at the low speeds that a hovering aircraft operates at, and have a number of other problems that don't make them suitable for most applications outside of turbofans on high speed fixed wing aircraft. Unfortunately with aviation, everything is a tradeoff, and helicopters are still used for the applications they're used for because they're still the best at what they do.
This is one example. Venture capital has been buying in for a while and there are several companies competing to be first to develop one that meets FAA standards. You might be right that it won't...
This is one example. Venture capital has been buying in for a while and there are several companies competing to be first to develop one that meets FAA standards.
You might be right that it won't happen but they seem to be proposing the equivalent of a shuttle bus service.
VC, companies, and countries bought into the Hyperloop which was obviously fraudulent and stupid out of the gate. I believe air taxi's are an order of magnitude more likely, but I also don't put...
VC, companies, and countries bought into the Hyperloop which was obviously fraudulent and stupid out of the gate. I believe air taxi's are an order of magnitude more likely, but I also don't put any stock by companies shoveling money around. Especially when deeper investigation off reveals all sorts of shenanigans in how those deals are brokered, handled, how much money was actually given, and what other incentives (like tax breaks) caused the investment.
As for Joby, the only thing that's moderately interesting to me from their own press releases is the supposed silence of these devices. I'm still very skeptical of the claims, and even the footage, since it's allllll straight from them, and is maybe contradictory.
For example we have this from your link:
In 2022, Joby collaborated with NASA to measure the sound of its aircraft, confirming it registered the equivalent of 45.2 A-weighted decibels (dBA) when flying overhead at an altitude of 1640 feet (500 meters) – quieter than a typical conversation.
which is a rather odd metric, but ok (500 meters seems arbitrary in some ways but ok fine).
Db is a logrithmic scale so the math gets wonky, and both of these are press release tests so god knows there's a lot of variables not being controlled for, but i'm super suspicious of these numbers and the actual details.
If it can however do that, and come in anywhere near the cost of a helicopter, then yes it might actually have a niche. However for reference, a helicopter is $500,000+ machine. A bus can be too(depends on a lot of things), but a bus can hold 30-100 people, while these can hold 4.
What this means is that, again, I don't see this being something the normal person will EVER use, or at least outside of their work. Large sites like ports, military bases, airports, industrial sites, etc might have a use for it, and the wealthy/corporate kind might use it, but this isn't likely to change the world much more than "oh there's less helicopters".
It also doesn't help that:
Joby had a crash in the last 3 years.
Joby, like MANY of these snake oil salesman taking after musk, is one of the companies claiming fully autonomous aircraft.
If this is only marketing where in reality it's just yet another pilot and yes you can "automate" certain scenarios (while still being monitored by a pilot, ground control, a zillion sensors....), but if they're remotely serious then that's another massive red flag. These things are not simple and as much as the marketers love to prey on the naive understanding most people have, we are no where near that kind of automation.
Well they've already started replacing mental health workers, which I would argue potentially has more harm than a robot muffing up a bandaid. *Eye roll They need to stop it with automating high...
Well they've already started replacing mental health workers, which I would argue potentially has more harm than a robot muffing up a bandaid. *Eye roll
They need to stop it with automating high value jobs. But of course thats where the money is
I know you know this, so I'm commenting rhetorically, but The "they" making the choice are never the ones who suffer the consequences, but are always the ones who reap the reward under capitalism....
I know you know this, so I'm commenting rhetorically, but
They need to stop it with automating high value jobs.
The "they" making the choice are never the ones who suffer the consequences, but are always the ones who reap the reward under capitalism.
If our economic system doesn't reward high value jobs, and for many jobs it doesn't (educators), they will get replaced by a minimally marketable version. And the difference in value between the original and the replacement will go into the bank accounts of business owners, benefitting fewer people. In the case of fully consolidated industries, the number of people benefitting is even tinier.
It's been a while since I've been in this kind of robotics space, but I'd say the author's assertions generally align with what I've seen. Fundamentally, the dexterity tasks are very complex in a...
It's been a while since I've been in this kind of robotics space, but I'd say the author's assertions generally align with what I've seen. Fundamentally, the dexterity tasks are very complex in a highly variable environment that requires a lot of classification and inference to interact with in a generalized way. For example, given an array of objects on a table, the robot has to try to identify them, then develop a strategy based on the task and the material. If you grip a ball of putty the way you grip a cue ball, you'll crush it. If you grip a cue ball the way you grip a rubber ball, it will probably slip in the grip.
Realizing that this was over ten years ago now, I had a small role in the DARPA Rescue Challenge. I got to see a lot of the development around manipulation. They had a software tool that let you place a fixture in the world model, and the robot would plan a path to put the manipulator in that position. That was a good combination of human and robotic teaming, which is something CMU/NREC was known for. But as you can see, it was still very slow. CHIMP only came in third place, but they also nerfed the competition, and if they had kept some of the harder tasks in, I think it would have won.
The coolest thing about the video I linked is around 1:00, the robot falls over, rights itself, and continues. Although it was not in the scoring rubric, I think a real(ish) world demonstration of that kind of resilience was the biggest success of the competition (but also I am definitely biased).
The video, CHIMP was 6 year ago? I'm impressed and kind of scared. The task with cutting drywall: why is there a bunch of power tools, are they slightly different or just if they drop one pick up...
The video, CHIMP was 6 year ago? I'm impressed and kind of scared. The task with cutting drywall: why is there a bunch of power tools, are they slightly different or just if they drop one pick up another? It seems like the three finger grip works well enough for medium sized objects, without going full human hand level of joints .
This is why I wasn't very impressed with Beijing's robo half marathon: they didn't get back up by themselves to finish the much simpler task of following a clearly defined path. And that was a get around assisted race, not dexterity race.
The posted date in the video is wrong (somebody must have reposted it). The timestamp in the video (2015) is likely correct. The program started 15 months further back. This competition was the...
The posted date in the video is wrong (somebody must have reposted it). The timestamp in the video (2015) is likely correct. The program started 15 months further back. This competition was the end of the program. This robot was built from scratch, hardware and soft, in that time. Truly a monumental effort.
I think the idea with the different tools is that the program designers wanted there to be a variety of options to encourage creativity. The manipulator on CHIMP was a COTS item and did well enough, though I'm sure given more time we would have developed our own.
Not surprising that CHIMP blew through the drywall challenge. The original competition called for the robot to climb a ladder, so the limb motors are specced to allow the robot to lift itself (it weighs around 400lbs).
Chimp is a pretty amazing design, really focused on the task rather than being a general purpose robot. It is statically stable in the two track or four track configuration, and it can traverse a pretty crazy terrain in the four track mode. The debris clearing at the end was originally a pile of CMU blocks that the robots were expected to climb over.
There were other aspects of the challenge that are not readily apparent from the video, like intermittent periods of severely restricted bandwidth to encourage onboard semiautonomous capability.
It's very very impressive!! I've seen shorts of a snake rescue bot, and a bunch of those Big Dog videos but yeah handling tools is really cool. Was it driving its own car at the start?? If this...
It's very very impressive!! I've seen shorts of a snake rescue bot, and a bunch of those Big Dog videos but yeah handling tools is really cool. Was it driving its own car at the start?? If this was 2015....where are we (humanity we) at with this kind of hot these days?
It was indeed driving its own car. This challenge was set up in the wake of the Fukushima nuclear disaster. The general shape of the challenge was to build technology that could do rescue work...
It was indeed driving its own car. This challenge was set up in the wake of the Fukushima nuclear disaster. The general shape of the challenge was to build technology that could do rescue work with equipment and infrastructure designed for humans in an environment that wasn't safe for humans.
Can confirm. Robot dexterity is still very hard. The problem comes from a few different avenues: humans can bend, our bones can bend, our 'range' of motion in proportion to our muscular strength...
Exemplary
Can confirm. Robot dexterity is still very hard.
The problem comes from a few different avenues: humans can bend, our bones can bend, our 'range' of motion in proportion to our muscular strength is significant and kind of fucked up to be able to do with any sort of success.
Part of this is because we have a high-amount of sensational touch in our digits, our fingers, and our general body. Due to the amount we can touch, we can discern the appropriate way to adjust our grip, how to grip, and how much pressure and easily modulate at a fraction of a millimeter our ability to touch. Part of this is due to the way our muscle systems work - we tend to think of them as solid masses, but they are groupings of smaller muscles, which are grouping of even smaller, with incredibly accurate sensors (nerves) embedded inside each of them at the center - not only that but the muscles themselves are sensors.
In robots we just... can't do that. We have to have layerings. You have a membrane, then a sensor, then a membrane. Or a motor and a sensor. Not a solid state combination pizza hut and taco bell sort of situation.
The muscle both works in tandem with others, and each muscle has 0.00000X% of the weight distribution of the whole, but has 0.00X% of the strength in distribution, and each one of those have a series of nerves to modulate pressure and resources (shifting of weight) while being a sensor in itself. This is why the upper class fears the lower class gaining collective class consciousness again, as together we are strong and mighty, but individually we are weak to hold ourselves against the crushing weight of greed and oppression.
Whereas with robots, you have a rather bulky 'central' motor, and/or linkages with external sensors that do not and can not have the incredibly minuscule amount of shifting and precision needed to adjust to all varieties of objects.
Additionally, we as humans, we as animals, have had a lifetime to learn the intricacies of how to hold things in different ways, how to modulate our pressure, how to go from 0lbs of pressure to 60lbs in a matter of seconds in only a fraction of our overall dexterity mass. Robots don't have that yet, and if they did it would be expensive and likely break, a lot, being even more expensive.
Humans are malleable, we break things all the time. Sometimes a full on 'snap your bone' break, and sometimes in a more 'over-extended' when lifting something, or even cut yourself because you lifted something on a sharp edge that weighed just a hair too much. But, humans are also fairly good at 1) re-allocating our systems to repair the damage 2) become stronger from breaking things (like exercise, lifting weights etc) from exposure and 3) cheaper on the whole to repair broken bones and muscles by relaxing and eating and drinking something with electrolytes, like Brawndo, instead of heavy (or incredibly light) expensive custom made components that do not self-heal.
Then you get into a whole other component: Our brain doesn't even know how to do a lot of the muscle systems. It has no idea. Our muscles have memory (not how we used to think, the human body literally stores memories outside the brain, in organs, cells, and even bones)and can operate without the brain working with a million instances of sensor and pressure information, because its localized nervous system allows it to activate repeatable known movements, modulate weight and tensions, without having to converse with the CPU of the human system, the brain at all times back and forth, like we do with robots at the moment.
We just don't have the materials, the size, the self-healing, or even the knowledge of how to hold things, especially all things, and the general ability to reproduce human dexterity. Most humans don't even consciously know how to walk - and that's why robots were thought to never reproduce human walking in the 1980s and had reached a dead-end, to never be solved, because we were trying to train them to walk like we think humans, and ourselves walk. Do you know what was one of the biggest (imo) breakthroughs in robotic locomotion? It was in 1989. Ten years in to the 'unsolvable problem' of human-like locomotion without a tail. Bipedal, no counter weight movement. It wasn't sashaying and shaking those hips.
It was falling.
Humans don't walk like animals do with tails, they fall. We do control falling. We take a step, we fall forward, and we catch ourselves.
This one thought changed overnight the entire 'problem' of walking, but we needed the materials that could handle the stress of controlled falling over and over, and we needed the ability to have heightened sensors to do controlled falling. In one night we solved bipedal movement AND we were able to fall up the stairs. Now anyone can do it, but in terms of dexterity - it's a much harder, more required problem, requiring far more maths, faster calculations, more individualized knowledge for every type of holding, insane levels of minute modulation of pressure, and faster, smaller, cheaper, more embedded sensors.
We're in a similar spot right now. We can do directed, controlled, large scale 'monolithic' control systems, but we just can't quite replicate the human ability yet to have our motors be in the thousands in a single hand and be a self-regulating hivemind-esque collective-individual sensor-control system.
That is until next year, when I absolutely have the solution like everyone else has thought, so if anyone have a few billion laying around, I promise it will work. Maybe. Possibly. /s... unless?
Thanks for responding so thoroughly, and providing all the insights! cc: @boxer_dogs_dance, in case you missed this, since you were the one that initially asked the question. :)
Thanks for responding so thoroughly, and providing all the insights!
cc: @boxer_dogs_dance, in case you missed this, since you were the one that initially asked the question. :)
Could we sooner see overseas outsourcing of home chores? It'd be so much cheaper to have a human robotically fold my laundry remotely than to sponsor them with a visa. They can earn foreign currency without leaving their loved ones. A sort of Da Vinci surgical system used to scramble eggs and put groceries away. Use the basic robot to get around the house, have the remote operator take over the hands to pass the dexterity check.
Dude. I would say most people would be incapable of assembling or replacing the escapement assembly even after watching a few demos, without downright wrecking the watch or at the very least they might be able to cram it in but not balancing it properly. Without training or existing familiarity I would be very willing to bet almost no humans would be able to do this. Replace a quartz watch battery is a more fair task to ask of regular humans.
We already have any number of similar machines that roll dough into precise smaller quantities. This one sounds easier if evals involve accessing a bigger database of what other robots know....right?
There is actually a pretty good scifi movie based on that premise called Sleep Dealer (Trailer):
https://en.wikipedia.org/wiki/Sleep_Dealer
I really wonder how a system like this would impact the US politically. Lots of blue collar manufacturing jobs have already been exported to countries where labor is cheaper since the cost to ship the goods back is low. This has caused a ton of discontent with US citizens who would otherwise work in domestic manufacturing.
Construction and the service industry is now the best option for many blue collar workers in the US who want to make a decent living without a college education. If we use remotely controlled robots to fill those positions for less money, what is left for US workers?
I know UBI is the logical conclusion, but that will likely come after mass discontent/unrest because I don’t count on US legislators to be proactive about anything. Ideally, blue collar workers would just upskill and get a higher education, but not everyone wants to or is able to get a college degree.
I personally think that if you want to do well in society and society changes, you should try to adapt to the new reality. I also understand that when a small group of people refuses to or is unable to adapt to a changing economy, it’s generally their personal problem to deal with. When a significant percentage of the country struggles to deal with changes in the economy it becomes the government’s problem.
I am with you. As much as I love watching those Boston Dynamics videos of robots jumping around in construction sights, where is my cleaning robot or massage robot?!?! This is going to be like the flying cars hopes and dreams of this generation. I want Rosey from the Jetsons, damn it.
I guess for me I learned a while ago that machines and automation, are not always for the machine to do something "better" or even at the same level of a person. Sometimes it is a machine that works good enough, to let the person do other things. Sometimes its to reduce the hazards associated with a human doing it (bomb diffuser, or repetitive tasks). I am mostly speaking about robots I used in labs before, they were not faster than a human doing the thing, but they allowed the humans to go do other things that only the humans could do. Automated machines are also generally more consistently precise than humans, so they are a good use case there; additionally, robots don't need sleep allowing them to perform things at a scale that a human cannot, but they don't do the individual thing better, necessarily.
Think about roombas. For me the amount of effort to make sure my floor is spotless so that the Roomba can operate without my attention, is so much work, at that point I'd rather clean the floor myself. It's like people who developed the roomba never cleaned in their lives, or were just the final person who mops a completely cleared out floor..."wow cleaning is so easy".
There's a reason automation and robots have not taken over the world, yet, and it's because at the end of the day they are not affordable over time (who maintains the robots and how expensive are those people, how replaceable are those people) and/or they don't do the job well enough (cleaning). Additionally, many people will literally do work for free. There is nothing cheaper than work for free.
Nobody does because Roombas suck. The Roborock Saros Z70 is where it's at. It actually does the thing you're talking about. It has an arm and picks up stuff from the floor. It's not that they didn't think about it when designing a robot vacuum, it's just very hard.
Even one without an arm is still a net benefit. It allows me to not do the vacuuming and mopping, but it does force me to pick stuff up and put it away.
It ties into your larger point, robots should exist to reduce human effort and my robot vacuum definitely does.
I somewhat hate these demos, because they set the wrong expectations. What they're showing off IS fascinating as hell, but every single move has been preprogrammed to an absurd degree. It's not "just run over there" or "do a dance". It's a ton of very precise code telling it how to move its arms/legs/etc to get those outputs.
It's still VASTLY less code than literally hard coding it (which still would've been impossible up until recently), but the idea that these things are anywhere close to generic tasks, let alone cheaply, is 100% in flying car territory.
Both in the lack of tech, and likely the lack of practicality. There's a reason you don't have a stationary clothes folding maching in your house, even though factories have had them for decades. It's big, loud, and expensive. Only worth it if you're folding THOUSANDS of shirts, not your 20-100.
Are you distinguishing flying cars from air taxis?
There are several companies nearly ready to offer air taxi service to the public although I don't know their price point.
Yes, although it somewhat depends on your definitions of air taxi, given that's something that has existed for some time for planes and helicopters.
If you mean the quad copter style things, that's basically just a helicopter with different steps (maybe arguably better given how hellishly complicated helicopters are), and I don't think comes close to qualifying as something you can drive anywhere or use for point to point (if anything more equivalent to a bus or trolley).
Even more so, i sincerely doubt any of them are going to actually start any sort of service. The whole problem with ANY sort of flying car is just how insanely dangerous it is. Imagine if every single fender bender was fatal, that's the reality you're looking at. So of course we've built a MASSIVE system to make sure that everything in the air is super far apart, logs its EXACT path, follows explicit rules, and is constantly monitored by machines and humans both on the ground and in the air, and people still die.
In fact the most dangerous flying device you can get in is a helicopter. They fly low enough to the ground and have awkward enough aerodynamics that should something go wrong you have an insanely small window to HOPEFULLY fix it. Nothing about the new "air taxi's" i've seen addresses ANY of this.
The only reason i'm not certain is we have "self driving cars" that ALSO shouldn't be allowed given their lack of data, testing, and quality, and I see these air taxi's as a similar level of "90% hype with no thought for the actual legalities and issues this might cause", but apparently that's ok now.
If you've got an article that's got something saying otherwise I'm all ears, but everything I've seen sure as hell reminds me of https://en.wikipedia.org/wiki/Moller_M400_Skycar. Every air taxi copter thing i'm aware of only has the "Experimental" approval from the FAA, which basically means "not even remotely safe or close to it". Should they get any sort of actual approval from the FAA i'll keep my eyes open, but no one is "nearly ready" if they don't already have that.
They're kind of worse in virtually every way actually, aside from the complexity thing. They're less safe, because they lack the ability to autorotate, so if you have a total power loss, not only do you just drop like a rock, you have zero control authority over the aircraft, because motors are the only thing giving you control authority, and they require power. They're less efficient because they use small props instead of large rotors. They're louder, because the blades need to spin much quicker to generate the same amount of lift, and there are more of them, and then obviously they have extremely limited flight time due to being electrically powered. There's a reason multirotors haven't supplanted helicopters for human transport in general. Those reasons haven't really changed just because you're charging fares for shorter trips now.
I really don't see these things ever taking off.
Yeah that was my expectation. Digging into the one linked it's supposedly somehow quieter, but it also appears to be using double blades per "wing" and also doing a V22 osprey thing where in flight they rotate and it flies "forward".
Granted sharing anything with the V22 also doesn't bode well.
It's likely because it uses ducted fans, which are quieter than props/rotors, but are also way less efficient at the low speeds that a hovering aircraft operates at, and have a number of other problems that don't make them suitable for most applications outside of turbofans on high speed fixed wing aircraft. Unfortunately with aviation, everything is a tradeoff, and helicopters are still used for the applications they're used for because they're still the best at what they do.
This is one example. Venture capital has been buying in for a while and there are several companies competing to be first to develop one that meets FAA standards.
You might be right that it won't happen but they seem to be proposing the equivalent of a shuttle bus service.
Toyota has bought in
Australia
VC, companies, and countries bought into the Hyperloop which was obviously fraudulent and stupid out of the gate. I believe air taxi's are an order of magnitude more likely, but I also don't put any stock by companies shoveling money around. Especially when deeper investigation off reveals all sorts of shenanigans in how those deals are brokered, handled, how much money was actually given, and what other incentives (like tax breaks) caused the investment.
As for Joby, the only thing that's moderately interesting to me from their own press releases is the supposed silence of these devices. I'm still very skeptical of the claims, and even the footage, since it's allllll straight from them, and is maybe contradictory.
For example we have this from your link:
which is a rather odd metric, but ok (500 meters seems arbitrary in some ways but ok fine).
However they also have this video https://www.youtube.com/watch?v=GHmXR0wBOiI which claims it's 55 db when only 100 meters away.
Db is a logrithmic scale so the math gets wonky, and both of these are press release tests so god knows there's a lot of variables not being controlled for, but i'm super suspicious of these numbers and the actual details.
If it can however do that, and come in anywhere near the cost of a helicopter, then yes it might actually have a niche. However for reference, a helicopter is $500,000+ machine. A bus can be too(depends on a lot of things), but a bus can hold 30-100 people, while these can hold 4.
What this means is that, again, I don't see this being something the normal person will EVER use, or at least outside of their work. Large sites like ports, military bases, airports, industrial sites, etc might have a use for it, and the wealthy/corporate kind might use it, but this isn't likely to change the world much more than "oh there's less helicopters".
It also doesn't help that:
If this is only marketing where in reality it's just yet another pilot and yes you can "automate" certain scenarios (while still being monitored by a pilot, ground control, a zillion sensors....), but if they're remotely serious then that's another massive red flag. These things are not simple and as much as the marketers love to prey on the naive understanding most people have, we are no where near that kind of automation.
I was most interested in the open package and apply a bandaid example since industry is hyping the possibility of replace health care workers.
Well they've already started replacing mental health workers, which I would argue potentially has more harm than a robot muffing up a bandaid. *Eye roll
They need to stop it with automating high value jobs. But of course thats where the money is
I know you know this, so I'm commenting rhetorically, but
The "they" making the choice are never the ones who suffer the consequences, but are always the ones who reap the reward under capitalism.
If our economic system doesn't reward high value jobs, and for many jobs it doesn't (educators), they will get replaced by a minimally marketable version. And the difference in value between the original and the replacement will go into the bank accounts of business owners, benefitting fewer people. In the case of fully consolidated industries, the number of people benefitting is even tinier.
It's been a while since I've been in this kind of robotics space, but I'd say the author's assertions generally align with what I've seen. Fundamentally, the dexterity tasks are very complex in a highly variable environment that requires a lot of classification and inference to interact with in a generalized way. For example, given an array of objects on a table, the robot has to try to identify them, then develop a strategy based on the task and the material. If you grip a ball of putty the way you grip a cue ball, you'll crush it. If you grip a cue ball the way you grip a rubber ball, it will probably slip in the grip.
Realizing that this was over ten years ago now, I had a small role in the DARPA Rescue Challenge. I got to see a lot of the development around manipulation. They had a software tool that let you place a fixture in the world model, and the robot would plan a path to put the manipulator in that position. That was a good combination of human and robotic teaming, which is something CMU/NREC was known for. But as you can see, it was still very slow. CHIMP only came in third place, but they also nerfed the competition, and if they had kept some of the harder tasks in, I think it would have won.
The coolest thing about the video I linked is around 1:00, the robot falls over, rights itself, and continues. Although it was not in the scoring rubric, I think a real(ish) world demonstration of that kind of resilience was the biggest success of the competition (but also I am definitely biased).
The video, CHIMP was 6 year ago? I'm impressed and kind of scared. The task with cutting drywall: why is there a bunch of power tools, are they slightly different or just if they drop one pick up another? It seems like the three finger grip works well enough for medium sized objects, without going full human hand level of joints .
This is why I wasn't very impressed with Beijing's robo half marathon: they didn't get back up by themselves to finish the much simpler task of following a clearly defined path. And that was a get around assisted race, not dexterity race.
The posted date in the video is wrong (somebody must have reposted it). The timestamp in the video (2015) is likely correct. The program started 15 months further back. This competition was the end of the program. This robot was built from scratch, hardware and soft, in that time. Truly a monumental effort.
I think the idea with the different tools is that the program designers wanted there to be a variety of options to encourage creativity. The manipulator on CHIMP was a COTS item and did well enough, though I'm sure given more time we would have developed our own.
Not surprising that CHIMP blew through the drywall challenge. The original competition called for the robot to climb a ladder, so the limb motors are specced to allow the robot to lift itself (it weighs around 400lbs).
Chimp is a pretty amazing design, really focused on the task rather than being a general purpose robot. It is statically stable in the two track or four track configuration, and it can traverse a pretty crazy terrain in the four track mode. The debris clearing at the end was originally a pile of CMU blocks that the robots were expected to climb over.
There were other aspects of the challenge that are not readily apparent from the video, like intermittent periods of severely restricted bandwidth to encourage onboard semiautonomous capability.
I also found a link to the journal paper, which has a lot more details.
It's very very impressive!! I've seen shorts of a snake rescue bot, and a bunch of those Big Dog videos but yeah handling tools is really cool. Was it driving its own car at the start?? If this was 2015....where are we (humanity we) at with this kind of hot these days?
It was indeed driving its own car. This challenge was set up in the wake of the Fukushima nuclear disaster. The general shape of the challenge was to build technology that could do rescue work with equipment and infrastructure designed for humans in an environment that wasn't safe for humans.
I would love it if experts weighed in on this article.
@drannex is our resident robotics expert, so they may be able/willing to offer you their opinion on the article. :)
Can confirm. Robot dexterity is still very hard.
The problem comes from a few different avenues: humans can bend, our bones can bend, our 'range' of motion in proportion to our muscular strength is significant and kind of fucked up to be able to do with any sort of success.
Part of this is because we have a high-amount of sensational touch in our digits, our fingers, and our general body. Due to the amount we can touch, we can discern the appropriate way to adjust our grip, how to grip, and how much pressure and easily modulate at a fraction of a millimeter our ability to touch. Part of this is due to the way our muscle systems work - we tend to think of them as solid masses, but they are groupings of smaller muscles, which are grouping of even smaller, with incredibly accurate sensors (nerves) embedded inside each of them at the center - not only that but the muscles themselves are sensors.
In robots we just... can't do that. We have to have layerings. You have a membrane, then a sensor, then a membrane. Or a motor and a sensor. Not a solid state combination pizza hut and taco bell sort of situation.
The muscle both works in tandem with others, and each muscle has 0.00000X% of the weight distribution of the whole, but has 0.00X% of the strength in distribution, and each one of those have a series of nerves to modulate pressure and resources (shifting of weight) while being a sensor in itself. This is why the upper class fears the lower class gaining collective class consciousness again, as together we are strong and mighty, but individually we are weak to hold ourselves against the crushing weight of greed and oppression.
Whereas with robots, you have a rather bulky 'central' motor, and/or linkages with external sensors that do not and can not have the incredibly minuscule amount of shifting and precision needed to adjust to all varieties of objects.
Additionally, we as humans, we as animals, have had a lifetime to learn the intricacies of how to hold things in different ways, how to modulate our pressure, how to go from 0lbs of pressure to 60lbs in a matter of seconds in only a fraction of our overall dexterity mass. Robots don't have that yet, and if they did it would be expensive and likely break, a lot, being even more expensive.
Humans are malleable, we break things all the time. Sometimes a full on 'snap your bone' break, and sometimes in a more 'over-extended' when lifting something, or even cut yourself because you lifted something on a sharp edge that weighed just a hair too much. But, humans are also fairly good at 1) re-allocating our systems to repair the damage 2) become stronger from breaking things (like exercise, lifting weights etc) from exposure and 3) cheaper on the whole to repair broken bones and muscles by relaxing and eating and drinking something with electrolytes, like Brawndo, instead of heavy (or incredibly light) expensive custom made components that do not self-heal.
Then you get into a whole other component: Our brain doesn't even know how to do a lot of the muscle systems. It has no idea. Our muscles have memory (not how we used to think, the human body literally stores memories outside the brain, in organs, cells, and even bones)and can operate without the brain working with a million instances of sensor and pressure information, because its localized nervous system allows it to activate repeatable known movements, modulate weight and tensions, without having to converse with the CPU of the human system, the brain at all times back and forth, like we do with robots at the moment.
We just don't have the materials, the size, the self-healing, or even the knowledge of how to hold things, especially all things, and the general ability to reproduce human dexterity. Most humans don't even consciously know how to walk - and that's why robots were thought to never reproduce human walking in the 1980s and had reached a dead-end, to never be solved, because we were trying to train them to walk like we think humans, and ourselves walk. Do you know what was one of the biggest (imo) breakthroughs in robotic locomotion? It was in 1989. Ten years in to the 'unsolvable problem' of human-like locomotion without a tail. Bipedal, no counter weight movement. It wasn't sashaying and shaking those hips.
It was falling.
Humans don't walk like animals do with tails, they fall. We do control falling. We take a step, we fall forward, and we catch ourselves.
This one thought changed overnight the entire 'problem' of walking, but we needed the materials that could handle the stress of controlled falling over and over, and we needed the ability to have heightened sensors to do controlled falling. In one night we solved bipedal movement AND we were able to fall up the stairs. Now anyone can do it, but in terms of dexterity - it's a much harder, more required problem, requiring far more maths, faster calculations, more individualized knowledge for every type of holding, insane levels of minute modulation of pressure, and faster, smaller, cheaper, more embedded sensors.
We're in a similar spot right now. We can do directed, controlled, large scale 'monolithic' control systems, but we just can't quite replicate the human ability yet to have our motors be in the thousands in a single hand and be a self-regulating hivemind-esque collective-individual sensor-control system.
That is until next year, when I absolutely have the solution like everyone else has thought, so if anyone have a few billion laying around, I promise it will work. Maybe. Possibly. /s... unless?
Weird tangent I know, I hope it makes sense.
Thanks for responding so thoroughly, and providing all the insights!
cc: @boxer_dogs_dance, in case you missed this, since you were the one that initially asked the question. :)