Complete aside here: I used to do work with amputees and prosthetics. There is a standardized test (and I just cannot remember the name) that fits in a briefcase. It's used for measuring the level of damage to the upper limbs and for prosthetic grading.
Basically, it's got the dumbest and simplest things in it. Stuff like a lock and key, a glass of water and jug, common units of currency, a zipper, etc. It tests if you can do any of those common human tasks. Like pouring a glass of water, picking up coins from a flat surface (I chew off my nails so even an able person like me fails that), zip up a jacket, lock your own door, put on lipstick, etc.
We had hand prosthetics that could play Mozart at 5x speed on a baby grand, but could not pick up a silver dollar or zip a jacket even a little bit. To the patients, the hands were therefore about as useful as a metal hook (a common solution with amputees today, not just pirates!).
Again, a total aside here, but your comment just reminded me of that brown briefcase. Life, it turns out, is a lot more complex than we give it credit for. Even pouring the OJ can be, in rare cases, transcendent.
There's a lot of truth in this. I sometimes joke that robot benchmarks should focus on common household chores. Given a basket of mixed laundry, sort and fold everything into organized piles. Load a dishwasher given a sink and counters overflowing with dishes piled up haphazardly. Clean a bedroom that kids have trashed. We do these tasks almost without thinking, but the unstructured nature presents challenges for robots.
Slightly tangential, we already have amazing laundry robots. They are called washing and drying machines. We don't give these marvels enough credit, mostly because they aren't shaped like humans.
Humanoid robots are mostly a waste of time. Task-shaped robots are much easier to design, build, and maintain... and are more reliable. Some of the things you mention might needs humanoid versatility (loading the dishwasher), others would be far better served by purpose-built robots (laundry sorting).
Sadly current "dishwasher" models are neither self-loading nor unloading. (Seems like they should be able to take a tray of dishes, sort them, load them, and stack them after cleaning.)
The problem is more doing it in sufficiently little space, and using little enough water and energy. Doing one that you just feed dishes individually and that immediate wash them and feed them to storage should be entirely viable, but it'd be wasteful, and it'd compete with people having multiple small drawer-style dishwashers, offering relatively little convenience over that.
It seems most people aren't willing to pay for multiple dishwashers - even multiple small ones or set aside enough space, and that places severe constraints on trying to do better.
There isn't a "task-shaped" robot for unstructured and complex manipulation, other than high DoF arms with vision and neural nets. For example, a machine which can cook food would be best solved with two robotic arms. However, these stationary arms would be wasted if they were just idling most of the time. So, you add locomotion and dynamic balancing with legs. And now these two arms can be used in 1000 different tasks, which makes them 1000x more valuable.
So, not only is the human form the only solution for many tasks, it's also a much cheaper solution considering the idle time of task-specific robots. You would need only a single humanoid robot for all tasks, instead of buying a different machine for each task. And instead of having to design and build a new machine for each task, you'll need to just download new software for each task.
I agree. I don’t know where this obsession comes from. Obsession with resembling as close to humans as possible. We’re so far from being perfect. If you need proof just look at your teeth. Yes, we’re relatively universal, but a screwdriver is more efficient at driving in screws that our fingers. So please, stop wasting time building perfect universal robots, build more purpose-build ones.
Given we have shaped so many tasks to fit our bodies, it will be a long time before a bot able to do a variety/majority of human tasks the human way won’t be valuable.
1000 machines specialized for 1000 tasks are great, but don’t deliver the same value as a single bot that can interchange with people flexibly.
The shape doesn't matter! Non-humanoid shapes give minir advantages on specific tasks but for a general robot you'll have a hard time finding a shape much more optimal than humanoid. And if you go with humanoid you have so much data available! Videos contain the information of which movements a robot should execude. Teleoperation is easy.
This is the bitter lesson! The shape doesn't matter, any shape will work with the right architecture, data and training!
Purpose build robots are basically solved. Dishwashers, laundry machines, assembly robots, etc. the moat is a general purpose robot that can do what a human can do.
Great examples. They are simple, reliable, efficient and effective. Far better than blindly copying what a human being does. Maybe there are equally clever ways of doing things like folding clothes.
I maintain that whoever invents a robust laundry folding robot will be a trillionaire. In that, I dump jumbled clean clothes straight from a dryer at it and out comes folded and sorted clothes (and those loner socks). I know we're getting close, but I also know we're not there yet.
We are certainly getting close! In 2010, watching PR2 fold some unseen towels is similar to watching paint dry [1], but we can now enjoy robots attain lazy student-level laundry folding in real-time, as demonstrated by π₀[2].
I can live without folding laundry (I can just shove my undershirts in the closet, who cares if it's not folded), but whoever manufactures a reliable auto-loading dishwasher will have my dollars. Like, just put all your dishes in the sink and let the machine handle them.
But if your dishwasher is empty is takes nearly the same amount of time/effort to put dishes straight into the dishwasher that it does to put them in the sink.
I think I'd only really save time by having a robot that could unload my dishwasher and put up the clean dishes.
That's called a second dishwasher: one is for taking out, the other for putting in. When the latter is full, turn it on, dirty dishes wait outside until the cycle finishes, when the dishwashers switch roles.
I thought about this and it gets even better. You do not really need shelves as you just use the clean dishwasher as the storage place. I honestly don’t know why this is not a thing in big or wealthy homes.
Hmm, that doesn't match my experience. It takes me a lot more time to put dishes into the dishwasher, because it has different places for cutlery, bowls, dishes, and so on, and of course the existing structure never matches my bowls' size perfectly so I have to play tetris or run it with only 2/3 filled (which will cause me to waste more time as I have to do dishes again sooner).
And that's before we get to bits of sticky rice left on bowls, which somehow dishwashers never scrape off clean. YMMV.
1. Get a set of dishes that does fit nicely together in the dishwasher.
2. Start with a cold prewash, preferably with a little powder in there too. This massively helps with stubborn stuff. This one is annoying though because you might have to come back and switch it on after the prewash. A good job for the robot butler.
Why can't dishwashers just be small, single-dish appliances in which you put the plate/mug/wine glass/forks/whatever, close it, push a button, and 10 seconds later it's clean and dry, you unload and repeat?
At this point I'm not sure we'll actually get a task-specific machine for laundry folding/sorting before humanoid robots gain the capability to do it well enough.
Foldimate has gone bankrupt in 2021 [1], and the ___domain referral from foldimate.com to a 404 page at miele.com, suggests that it was Miele who bought up the remains, not a sketchy company with a ".website" top-level ___domain.
Getting to LLMs that could talk to us turned out to be a lot easier than making something that could control even a robotic arm without precise programming, let alone a humanoid.
This was actually discovered quite early on in the history of AI:
> Rodney Brooks explains that, according to early AI research, intelligence was "best characterized as the things that highly educated male scientists found challenging", such as chess, symbolic integration, proving mathematical theorems and solving complicated word algebra problems. "The things that children of four or five years could do effortlessly, such as visually distinguishing between a coffee cup and a chair, or walking around on two legs, or finding their way from their bedroom to the living room were not thought of as activities requiring intelligence."
>I don't know why people always feel the need to gender these things
Because it's relevant to the point being made, i.e. that these tests reflect the biases and interests of the people who make them. This is true not just for AI tests, but intelligence test applied to humans. That Demis Hassabis, a chess player and video game designer, decided to test his machine on video games, Go and chess probably is not an accident.
The more interesting question is why people respond so apprehensively to pointing out a very obvious problem and bias in test design.
> i.e. that these tests reflect the biases and interests of the people who make them
Of course. However i believe we can't move past that without being honest about where these biases are coming from. Many things in our world are the result of gender bias, both subtle and overt. However, at least at first glance, this does not appear to be one of them, and statements like the grandparent's quote serve to perpetuate such biases further.
Then remove the parts that offend your modern sensibilities and focus on the essence.
He was right. Scientists were focusing on the "science-y" bits and completely missed the elephant in the room, that the thing a toddler already masters are the monster challenge for AI right now, before we even get into "meaning of life" type stuff.
I had a pretty bad case of tendinitis once, that basically made my thumb useless since using it would cause extreme pain. That test seems really good. I could use a computer keyboard without any issue, but putting a belt on or pouring water was impossible.
I had a swollen elbow a short while ago, and the amount of things I've never thought about that were affected by reduced elbow join mobility and an inability to put pressure on the elbow was disturbing.
It feels like there's a whole class of information that easily shorthanded, but really hard to explain to novices.
I think a lot about carpentry. From the outside, it's pretty easy: Just make the wood into the right shape and stick it together. But as one progresses, the intricacies become more apparent. Variations in the wood, the direction of the grain, the seasonal variations in thickness, joinery techniques that are durable but also time efficient.
The way this information connects is highly multisensory and multimodal. I now know which species of wood to use for which applications. This knowledge was hard won through many, many mistakes and trials that took place at my home, the hardware store, the lumberyard, on YouTube, from my neighbor Steve, and in books written by experts.
I think assembling Legos would be a cool robot benchmark: you need to parse the instructions, locate the pieces you need, pick them up, orient them, snap them to your current assembly, visually check if you achieved the desired state, repeat
I agree. Watching my toddler daughter build with small legos makes me understand how incredible fine motor skills are as even with small fingers some of the blocks are just too hard to snap together.
There's also the option of not pulling the pin, and shooting your enemies as they instinctively run from what they think is a live grenade. Saw it on a TV show the other day.
> We had hand prosthetics that could play Mozart at 5x speed on a baby grand, but could not pick up a silver dollar or zip a jacket even a little bit. "
I must be missing something, how can they be able to play Mozart at 5x speed with their prosthetics but not zip a jacket? They could press keys but not do tasks requiring feedback?
Or did you mean they used to play Mozart at 5x speed before they became amputees?
Playing a piano involves pushing down on the right keys with the right force at the right time, but that could be pre-programmed well before computers. The self-playing piano in the saloon in Westworld wasn't a huge anachronism, such things slightly overlapped with the Wild West era: https://en.wikipedia.org/wiki/Player_piano
Picking up a 1mm thick metal disk from a flat surface requires the user gives the exact time, place, and force, and I'm not even sure what considerations it needs for surface materials (e.g. slightly squishy fake skin) and/or tip shapes (e.g. fake nails).
Imagine a prosthetic 'hand' that has 5 regular fingers, rather than 4 fingers and a thumb. It would be able to play a piano just fine, but be unable to grasp anything small, like a zipper.
I'm far from a piano player, but I can definitely push piano buttons quite quickly while zipping up my jacket when it's cold and/or wet outside is really difficult.
Even more so for picking up coins from a flat surface.
For robotics, it's kind of obvious, speed is rarely an issue, so the "5x" part is almost trivial. And you can program the sequence quite easily, so that's also doable. Piano keys are big and obvious and an ergonomically designed interface meant to be relatively easy to press, ergo easy even for a prosthetic. A small coin on a flat surface is far from ergonomic.
I play piano as a hobby, and the funny thing is, if my hands are so cold that I can't zip up my jacket, there's no way I can play anything well. I know it's not quite zipping up jackets ;) but a human playing the piano does require a fast feedback loop.
But how do you deliberately control those fingers to actually play yourself what you have in mind rather than something preprogrammed? Surely the idea of a prosthetic does not just mean "a robot that is connected to your body", but something that the owner control with your mind.
Nobody said anything about deliberately controlling those fingers to play yourself. Clearly it's not something you do for the sake of the enjoyment of playing, but more likely a demonstration of the dexterity of the prosthesis and ability to program it for complex tasks.
The idea of a prosthesis is to help you regain functionality. If the best way of doing that is through automation, then it'd make little sense not to.
Well, you see, while the original comment says they could play at 5x speed, it does not say it plays at that speed well or play it beautifully. Any teacher or any student who learned piano for a while will tell you that this matters a lot, especially for classical music -- being able to accurately play at an even tempo with the correct dynamics and articulation is hard and is what differentiates a beginner/intermediate player from an advanced one. In fact, one mistake many students make is playing a piece too fast when they are not ready, and teachers really want students to practice very slowly.
My point is -- being able to zip a jacket is all about those subtle actions, and could actually be harder than "just" playing piano fast.
zipping up a jacket is really hard to do, and requires very precise movements and coordination between hands.
playing mozart is much more forgiving in terms of the number of different motions you have to make in different directions, the amount of pressure to apply, and even the black keys are much bigger than large sized zipper tongues.
Pretty much. The issue with zippers is that the fabric moves about in unpredictable ways. Piano playing was just movement programs. Zipping required (surprisingly) fast feedback. Also, gripping is somewhat tough compared to pressing.
Basically, it's got the dumbest and simplest things in it. Stuff like a lock and key, a glass of water and jug, common units of currency, a zipper, etc. It tests if you can do any of those common human tasks. Like pouring a glass of water, picking up coins from a flat surface (I chew off my nails so even an able person like me fails that), zip up a jacket, lock your own door, put on lipstick, etc.
We had hand prosthetics that could play Mozart at 5x speed on a baby grand, but could not pick up a silver dollar or zip a jacket even a little bit. To the patients, the hands were therefore about as useful as a metal hook (a common solution with amputees today, not just pirates!).
Again, a total aside here, but your comment just reminded me of that brown briefcase. Life, it turns out, is a lot more complex than we give it credit for. Even pouring the OJ can be, in rare cases, transcendent.