3
$\begingroup$

Most robots that exist in the world today are factory robots. Not particularly humanoid, except the arm+hand concept. When you design a robot to replace the human task of pushing a vacuum cleaner around the house, you don't design a robot that resembles a human in any way.

Roboticists today are designing all sorts of robots: with wheels, with tracks, with six legs, etc. So my question is: should we expect robots of a near-future setting to be humanoid? Would they be bipedal?

Of course, one thing sophisticated robots (not hoover robots and factory robots) need is adaptability. The human body-form can do a lot of different things. But so can the hexapedal form, for example.

Mother Nature already dealt with this question of what body-plan to use for Her most sophisticated beings, and she settled on humanoid for reasons I don't fully understand, discussed a bit in this question.

Would people in the near-future bother manufacturing humanoid/bipedal robots, or could we get any job done with other body-plans?

$\endgroup$
16
  • 6
    $\begingroup$ This has truly been done to death in fiction starting way way way back with the science fiction father of the topic Isaac Asimov, personally I think he vastly overstated the uncanny valley effect to fit his narrative purposes and that, if anything, taking other factors into account the trend and preference is much more likely to be the reverse of what he suggested for some personal household automata as time moves on, if not for those tailored to specific jobs, but I digress, are you really telling us you can't find suitable material on this to fit your needs and answer your questions? $\endgroup$
    – Pelinore
    Commented Feb 24, 2023 at 15:07
  • 4
    $\begingroup$ "for Her most sophisticated beings" 😲 really, you think so, by what measure? 😁 $\endgroup$
    – Pelinore
    Commented Feb 24, 2023 at 15:09
  • 2
    $\begingroup$ What @Pelinore said. The water flea has nearly 50% more genes in its genome than we do; does that make it more sophisticated? $\endgroup$
    – jdunlop
    Commented Feb 24, 2023 at 19:15
  • 4
    $\begingroup$ It's relevant that biology has a real problem making rotary joints - like wheels. Nature didn't design humans with legs necessarily because they're better but because that's what it could do. Nature also has a huge problem trying to change body plans. All mammals, reptiles, dinosaurs, birds, amphibians and a lot of fish have four limbs, a design inherited from ages ago. Although nature can tweak each limb, it finds it very difficult to add another one. $\endgroup$ Commented Feb 24, 2023 at 23:33
  • 2
    $\begingroup$ @RobbieGoodwin yup and notice how they still have the same number of legs as when they originated 480 million years ago $\endgroup$ Commented Feb 25, 2023 at 1:21

13 Answers 13

23
$\begingroup$

Form follows function

When identifying what you want your robot to do, you make a list of priorities. If "interacting with people" isn't on the list, then you don't put any effort into their appearance. Making something "humanoid" is an additional qualification that is really too expensive for most development teams.

Just bipedal walking is an expensive proposition. Trying to fit all of the necessary sensors into a human-shaped bulb of a head is an unnecessary challenge. Trying to make dexterous human hands is something our world is still working on.

The super-expensive Honda ASIMO was built basically as a PR stunt. After doing some dancing around on stage, it spent the rest of its days as a receptionist. It is now a museum piece.

Humanoid forms are actually a detraction to the natural functions of robots. The problem with humanoid robots is that humans can't help anthropomorphising them. We get nervous about being naked in front of them. You make a toaster with a human face, and pretty soon people are waving signs that say "free the toasters!"

So, overall, if you don't need to make something humanoid, then you won't.

OTOH, if you do need to make something humanoid, you will. We're entering the age where humans will need to work side-by-side with robots. Being able to predict a robot's behavior can be the difference between successfully building an automobile and getting your hand broken, so there is good reason to model robots on something we're familiar with.

I'm reminded of C3PO, who is a protocol droid. R2D2 doesn't need to interact with people, so it's ok for him to look like a fire plug. C3PO's function was to provide translation services and explain things to guests. Any hospitality robot will inevitably trend towards humanoid because it makes their job easier.

$\endgroup$
13
$\begingroup$

Versatility: A good reason for a humanoid robot is to use things designed for humans.

If you run a hotel, you could have a robot vacuum cleaner shaped like R2D2, a robot bellhop that's basically a luggage cart with mechnical arms and a motor, and a valet robot that's basically a tow truck with autopilot.

Or you could buy one humanoid robot that can use an ordinary vacuum cleaner, push an ordinary luggage cart, and drive the cars itself. All these devices were designed to be used by humanoids, so a humanoid shape is really handy.

$\endgroup$
3
  • 1
    $\begingroup$ That's essentially the argument Tesla is making for why Optimus is humanoid, and also why their cars only use cameras (roads are made for eyes, not lidars, radars, or ultrasound). $\endgroup$ Commented Feb 26, 2023 at 17:34
  • 1
    $\begingroup$ The catch with it, in particular for the hotel case, is that the humanoid robot will be more expensive than the simpler special-purpose ones, quite possibly more than 3× as expensive. But even if it's less than 3× the price, this may not pay off because although the humanoid can do all the three special tasks, it can't do them simultaneously. $\endgroup$ Commented Feb 27, 2023 at 20:18
  • 1
    $\begingroup$ @leftaroundabout Yes, the details matter. There's probably a scenario where one humanoid would be enough, like a remote power substation or something, where the robot is inert until Headquarters signals it to swap out a broken a mechanical component. $\endgroup$ Commented Feb 27, 2023 at 22:17
10
$\begingroup$

Most robots are designed for a quite specific range of tasks. So, factory robots for example, are usually intended to do a repetitive set of tasks. They are not intended to be arbitrarily capable. They are supposed to be doing things like putting a part on a device, weld it, take a photo or x-ray, spray some paint on, and various other things like so.

So a human-form robot, especially one that began to look substantially like a human, would require a specific purpose that required that shape and appearance.

Even tasks that currently have human adapted interfaces are unlikely to be filled by human-appearing robots. Consider the autopilot of a plane, or the computer-driven function on some new cars. (I'm deliberately avoiding the brand name.) There is no human-resembling body required. The effort to build one would actually be a lot of trouble. It's easier to redesign the interface to the computer than to build something to act like a human.

The only real opportunity for robots that strongly resemble humans would be where the resemblance was the purpose. Let's be delicate about that purposes. Let us say "companionship" might be the goal.

There are already a few types of "robot dog" available. It's not too far to imagine moving from that to "robot friends." Recall the movie AI, with robot children for childless couples.

There might be room for robot nannies and robot nurses. They might be made to look like humans because it was comforting to their clients.

Or, to avoid putting too fine a point on it, there might be robots that you call. At night. That also appears in the movie AI. Jude Law plays a robot called Gigolo Joe.

The resemblance might not be universally popular, however. There is the effect of the uncanny valley. Some people find things that look nearly human to be quite disconcerting.

=====

There have been comments to the effect that if you want a robot to do a bunch of things humans currently do, then it will start to look like a human.

No.

Even the example: It is much easier to design a robot to drive any car if you are not required to make it human shaped.

Consider a head with eyes on a neck that can turn and look at mirrors or look behind to reverse or check if it is safe to change lanes. This is massively restrictive and a lot of work. Much easier and much more productive to have cameras that can be mounted on any convenient location on the car. Even human-driven cars are doing this today, with such things as lane-assist and backup cameras.

Consider a human body plan for moving the controls. Two arms, two legs, a torso. Making the linkages to look like a human body plan, then having them find the controls, then having them work the controls, is way too much work. Much easier to have a few servo motors to actuate the controls.

Consider getting in and out of the car. Much easier to have an install/removal robot. That would look like a panel van with one or two "factory" robot arms that could open the side door, open the car door, stick the various parts of the driver robot in or remove them, fasten the cameras on, etc.

Indeed, such linkages could be constructed such that the drive seat was relatively unobstructed and still available for a human to ride in.

As well, layering on demands that a single robot do many human tasks in a human fashion is moving into the "I want a buddy to fetch my car for me, do the laundry, make toast, help the kids with homework, have a beer with me, etc. etc. etc." In other words, it's becoming companionship rather than the tasks.

Otherwise your toaster would already look humanoid.

$\endgroup$
8
  • 3
    $\begingroup$ Robot dogs are being made for military purposes because the leg system is very flexible to move on rough terrain - as there's a lot of flexibility about where to put down each leg on the ground, whereas a wheel or track system makes it necessary to touch all points between here and there. Wheeled robots can't step over landmines or debris. $\endgroup$ Commented Feb 24, 2023 at 23:37
  • 7
    $\begingroup$ Outside of aesthetics/entertainment, isn't the real use-case for a humanoid robot the one where you want a general-purpose robot that can do anything a human could do, and therefore needs a human-like body to approximate a human's skill set? $\endgroup$ Commented Feb 25, 2023 at 5:45
  • 1
    $\begingroup$ @JeremyFriesner yes, because while autopilot for a plane or a car specifically designed for that model will look nothing like human, robot designed to operate any plane or car, including vintage ones without any electronics inside, will have to interact with human-adapted controls. $\endgroup$ Commented Feb 25, 2023 at 7:26
  • $\begingroup$ I think we currently have very differently shaped robots because we don't only know how to make robots that are better than humans when we ultra-specialise them for one thing. Mainly because we still have no clue how to make a general-purpose AI that can learn to do everything a human can learn to do. Since making single-purpose robot "brains" is all we know how to do, we also specialise their physical forms for that single task, and people only buy them when they want that single task done (and in commercial setups, frequently also design their whole setup around that specific robot). (1/2) $\endgroup$
    – Ben
    Commented Feb 27, 2023 at 4:15
  • 1
    $\begingroup$ But if general-purpose robots become a thing, then business will probably want humanoid robots so they can replace their staff without also redesigning their entire operating environment. They will want a robot that can be a drop-in replacement for a human worker, who necessarily used tools and workspaces designed for humans. And then those robots can be repurposed for any other formerly-human job without changing infrastructure. It's general-purpose robots that will lead to wide deployment of humanoid ones, not special-purpose-that-requires-humanoid-shape. (2/2) $\endgroup$
    – Ben
    Commented Feb 27, 2023 at 4:23
8
$\begingroup$

Uncanny Valley

I think Boba Fit's answer pretty well covers the main reason why we might not see humanoid robots much, which is that robots are normally "built to task" but I wanted to elaborate on what I believe is an important ancillary reason we don't see humanoid robots: they creep us out, with the phenomena being called "uncanny valley".

In short, humans are pretty comfortable with machines. Cars, drones, etc, all perfectly reasonable, even if they are AI and fully automated. As things start to look more and more human, though, there is some real monkey-brain thinking that kicks in and kinda creeps us out, and the closer to human they look, the more creeped out we get. Things "don't move right" or "don't look right". (Horror movies get a lot of mileage out of this too, making humanoid horror entities move in weird ways or do something not quite human, which naturally creeps us out.)

The link I started with does pose some solutions to the problem, but it is a problem.

Plus, purpose-built robots are also just easier. Building wheeled or tracked delivery robots isn't nearly as difficult as building a humanoid robot! The humanoid could have some real advantages. Maybe it can get on a bus. It's easier to see and not trip over. But it's also going to be bigger, heavier, more inherently dangerous (stepping on your toes, falling over, etc) and enormously more complicated to develop. Mechanically, the human form is really complicated, versus a simple tracked vehicle.

$\endgroup$
11
  • 1
    $\begingroup$ At the risk of being a touch risqué 🥴 the continued financial success of the animatronic sex doll industry would beg to differ 😁 .. & those in robotics working on child friendly companions and teaching aids or on robotic companions / care provision for the elderly would also like a word I'm sure 🤗 .. few if any of them seem to put much value in the uncanny valley hypothesis (at least not in relation to their products, which they're spending money developing) when you look at what they're doing. $\endgroup$
    – Pelinore
    Commented Feb 24, 2023 at 16:04
  • 1
    $\begingroup$ But being a little more serious, take Honda's Asimo as an example, is their anything even slightly uncanny valley about it? does it make you feel even slightly uncomfortable? the honest answer for most who aren't on some kind of spectrum is going to be no, because the uncanny valley effect only really comes into play when something is so close to the real thing in appearance that it can initially fool us into thinking it is the real thing until we pay closer attention or interact with it a bit, humanoid robots that don't attempt to look entirely human simply don't produce the response at all. $\endgroup$
    – Pelinore
    Commented Feb 24, 2023 at 16:15
  • 1
    $\begingroup$ @Pelinore I think that's the thing -- the Asimo is so far removed from even being a functional humanoid that there's no uncanny valley effect. Imagine, however, if it moved a lot faster (i.e. was actually useful as a robot) and had all the necessary articulation and was able to perform normal human tasks at normal human speeds (or faster). I think you'd start getting into an uncanny valley effect, even if you kept the blank screen face and clear robotic appearance. The key to avoiding the effect is to keep the robot as far from human as possible, not just in face, but also shape and movement. $\endgroup$
    – JamieB
    Commented Feb 24, 2023 at 16:41
  • 3
    $\begingroup$ Oh no it's a very functional humanoid, for a given value of functional, it just doesn't look like a person, a humanoid body doesn't mean a robot has to appear human, it can still be obviously mechanical 🙂 $\endgroup$
    – Pelinore
    Commented Feb 24, 2023 at 17:05
  • 2
    $\begingroup$ @JamieB And that is exactly pelinore's point. It's very, very easy to make something humanoid without it being uncanny, because there's plenty of space to the left of the uncanny valley for the humanoid robots to occupy. $\endgroup$
    – somebody
    Commented Feb 28, 2023 at 6:06
5
$\begingroup$

It makes sense for robots to be humanoid if they're expected to be generalists that are interchangeable with humans in their roles

A humanoid robot can use things designed for humans. They can sit at human chairs, type on human keyboards, drive human cars, wash-up in human sinks, fire human guns - you get the idea. For any one of those roles you can design a better robot, but if you want a robot that fits seamlessly into human life, you need a humanoid robot.

So for a situation where humanoid robots are common, you need a situation where roles are expected to be taken either by humans or robots on a regular basis. If any role is only performed by robots then it'll be better to use peripherals on your robot than have it pick up a designed-for-humans tool.

Perhaps laws limiting robots taking human roles were introduced so now your starship has a maximum of 20% robot crew or something, or perhaps you have population issues that mean there aren't enough humans for the roles, or perhaps its just cheaper to pay one of the masses of humans desperate for food to do the role than pay for expensive robots, or perhaps your robots are considered sapient and have flexible and extensive rights over what jobs they take rather than being tools you can slap a hose attachment on. There are many options.

$\endgroup$
2
  • 1
    $\begingroup$ While I agree in general I believe you could end up with robot-only jobs that require humanoid robots because they must function within human space to do their jobs. The firefighting robot must open doors, climb ladders etc to get to the fire. The fire might be in an area with exotic door controls (say, a dementia ward--doors designed that the people living there can't figure out how to open them. They will be very non-standard but designed for human hands.) $\endgroup$ Commented Feb 27, 2023 at 3:10
  • $\begingroup$ It really, really doesn't make any sense to have a robot sit on a chair and type on a keyboard. Sure, perhaps it's good if it can use a keyboard for emergencies to type in a password or something, but for that any pencil-shaped tool extension would also suffice. $\endgroup$ Commented Feb 27, 2023 at 20:34
4
$\begingroup$

Nature hasn't settled on a humanoid form. There's only been a handful of humanoid species, and all but one are extinct. If anything, the ideal form is apparently crablike, or perhaps some variety of beetle.

One of the more successful robots in the DARPA Robotics Challenge was the JPL Robosimian, which was loosely based on an ape-like form factor (which also turned out a bit crab-like or beetle-like), a quadruped with four grasping/walking limbs. It also used common hardware for its limbs, arms and legs being largely interchangeable, and it could arrange itself to roll around on wheels.

$\endgroup$
2
$\begingroup$

AI relies on a model of the human brain.

Most AIs in the future are based off human models. Their minds inherently value a humanoid form. Efforts to insert them into other forms have resulted in robopsychosis where the dehumanization makes them go insane.

Efforts to design non human based robots have gone poorly, and few animals have been well suited to replacement.

As such, most advanced robots are built in a humanoid form so they can function well.

$\endgroup$
2
$\begingroup$

I haven't seen anyone state what I believe to be the primary reason you don't see many humanoid robots: It is vastly more difficult and costly to build a humanoid robot.

Most motion devices we have to work with are rotary motors, solenoids, rotary and linear servos, etc. These devices are not conducive to moving the human shape. So, if you want to do anything robotically, it is almost guaranteed to be easier and cheaper to do that thing in a non-human form.

Absent some exotic devices being researched, a humanoid robot will need to be powered by hydraulics or pneumatics. Each muscle needs to be powered by it's own cylinder and tubing. You can do some things with cable to reduce the number of components, but it's just a beast of a job.

Even the most advanced publicly known semi-humanoid robot in the world (Boston Dynamics 'Atlas'), which is the result of millions of dollars of work and research, can only lift about 15 pounds. And for getting any real work done, it is outclassed by their dog robot, which is orders of magnitude less productive than the kind of square cart robots in an Amazon warehouse.

I'm not saying that Boston Dynamics (or anyone else's) human and animal shaped robots don't have capabilities that other (boxy) robots don't have. I'm just saying they are very expensive and only make sense for relatively rare use cases.

If there were magic to make them available, quick and strong, I have no doubt that humanoid robots would be very popular. As there is no such magic, we tend to get what can be made effective for a given task at a price that businesses can afford.

$\endgroup$
1
  • 1
    $\begingroup$ Good points 👍 but there are artificial muscles though which are beginning to see more use in robotics so that's perhaps an issue of the past that may not be particularly valid going forward or for a fictional future where they are better and more commonly used, certainly a valid point for the near past real world though. $\endgroup$
    – Pelinore
    Commented Feb 26, 2023 at 7:06
2
$\begingroup$

Robot soldiers might look somewhat human so the enemy might waste ammunition on them that is ineffective.

Other answers have thoroughly covered other reasons for or against.

One more: if a large subculture is so anti-robot that they resort to sabotage, making their targets harder to distinguish from humans would be a good strategy.

$\endgroup$
1
$\begingroup$

There are three things that you need to take in to account:

Firstly, would the robot be performing a specific talk using purpose built tools. Or would it be a multi purpose robot sharing tools with its owner.

For example, a Roomba uses its own built in vacuum unit so it can be constructed in a very simple non-humanoid form. Whereas a cleaning robot that would be using an off the shelf vacuum cleaner that a human would use would require limbs capable of holding it, and a body capable of doing things like moving furniture or bending down to hoover under furniture.

The same with a robot to wash dishes, would it be a dishwasher on wheels, or would it need to stand at a sink and handle a scrubbing brush?

Would a robot soldier carry a rifle, and need to load clips into it, or would they just be a tank with a Gatling gun on the top?

Secondly, what kind of environment would the robot need to be in, would it need to be static or to move, and if it did need to move would this be once a day to get to a recharging station or would it be in constant motion.

Thirdly, how much computing power would it have to spare. Movement is complicated, and the more you need to move the more computing power is required. A robot with 6 legs would require more computing power to move them around than a robot with 4 legs, while a robot with wheels would require almost no computing power except that needed to navigate.

$\endgroup$
1
$\begingroup$

They wouldn't. Just look at nature and the real world.

If you've browsed the questions in this site about realistic angels and mermaids, you probably found answers that were a variation of "they're not realistic as they're depicted because the human body plan is anything but ideal for doing these". Humans are very similar from apes, they're our closest relatives, but it's nearly impossible to mistake an adult human and an adult chimpanzee, it's also nearly impossible for a human to beat a chimpanzee when it comes to move quickly by swinging through branches.

When you look at the types of animals in nature and the types of robots in the world today, you can easily arrive to the same conclusion: there's no such thing as a "one size fits all", and that's also true for shape. Something specialized from flying at high speeds will not be the best at ambushing something 5 km underwater. Similarly, a robot that's ideal for doing something in a factory line will look nothing like the one designed to film things from above. It makes no sense for humanoid robots to be built for everything because the humanoid shape is not good for everything. Many people on this site and many biologists will even agree that the humanoid form isn't even the best for humans, since the requirements to achieve what we've achieved are basically the ability to use tools and a brain like ours. For all intents and purposes a species of crow-like creatures with dexterous pincer-like mouthparts could probably achieve everything we have while also having to deal with a society of flying individuals.

Both in nature and in mechanics, one rule almost always applies: form follows function. Even in animals with traits that make it harder to survive, these traits usually play a role that ultimately makes them useful enough to be selected, like the feathers in male peacocks.

So would humanoid robots take over? No, not at all. Just looking at the real world shows us how much easier it is to make things that aren't meant to look, move and act like humans. There is however one thing a humanoid robot would be best as: substitute companions and caretakers.

Humans, at the end of the day, are social animals that have vastly modified the world and built machines to suit their needs and with their own form in mind. Based on this, a humanoid form is the ideal form with the function is to take care of and provide comfort to another human. The uncanny valley is a thing, but still humans are usually at their most comfortable when they're dealing with something that's also mostly human shaped. Being capable of doing everything a human can also enables these robots to perform nearly any function a normal human would need assistance with: be it cooking, taking them somewhere or operating other machines.

So summing it up: no, humanoid robots wouldn't take over as the one form of all machines, because that's just dumb. Modern engineering shows us how hard it is to make robots that can get even close to moving and acting like humans, with just about 0 current examples of one that can mostly move, talk and look like one. However, it's easy to make a drone or a roomba, and it's apparently also easier to make cars that drive themselves. The humanoid form is the best at dealing with and using equipment designed for humanoids. Other than that the human form is often not the best.

$\endgroup$
1
$\begingroup$

Obviously if people want general purpose servant robots to replace servants or slaves they would have to have humanoid body form to be able to perform many or all of the things a human servant could do.

And obviously sexbots would be built to look very human.

Some people might want robot companions (and maybe caretakers) which might look like dogs, or adult humans, or human children, depending on what sort of companions they wanted.

War robots are already being researched and developed, and mostly look like weird machines. But possibly some army might design general purpose war robots which have the same shape and size as human soldiers, and can be programmed to function as various types of soldiers, sailors, air crew, etc. with modules for different functions. Thus they can use weapons designed for human use and a country might build billions of them.

Even highly automated factories still have some human workers, and many third world factories have many human workers instead of automation. And I can imagine a process in which men, women, and children are replaced in factories by man, woman, or child sized humanoid robots which can be programmed to do any task a human can.

And possibly people in the future will not work but will own a few robot "slaves" they lease to various businesses and live on the rent paid for the use of their robot "slaves". An ordinary citizen might own a dozen, richer people might have sole or shared ownership of hundreds and/or invest in companies having many millions to rent out, and rich people might own millions of robots.

If married couples have fewer children in the future than the replacement number the population will decline. And possibly humanoid robots with human level intelligence may be created to perform the task formerly performed by humans, and to gradually replace the diminishing human population and live in their designed for humans structures, so that human civilization will not die out even when humans do.

And humanoid robot bodies might be built to be remotely controlled by humans who thus can avoid the unhealthy environments like deserts, the sea floor, outer space, etc. the robots operate in. If the humans want to experience being in those locations without actually being there, the robots should be highly humanoid and there should be methods for the human operators to see, hear, and feel though the sensory devices of the robots. And the robots might use tools and devices designed for humans to take advantage of the long practice the human operators might have with those tools and devices. Thus those robots controlled by telepresence for work or play might be built with humanoid bodies.

So I have listed a number of reasons why humanoid robots might conceivably be built in large numbers and be common in some hypothetical futures.

Many other answers have mentioned reasons why in many cases it would be more reasonable to build highly nonhumanoid robots designed for very specific single tasks. Most work robots today are highly non humanoid.

And that has been more or less predicted long before any working robots were built.

See, for example, "Q.U.R." by Anthony Boucher, Astounding Science Fiction March, 1943.

https://www.isfdb.org/cgi-bin/title.cgi?46709

$\endgroup$
0
$\begingroup$

Human-looking robots typically interact with humans more, as designed.

Humanoid shapes are meant for you (the meat bag) to relate to the robot, a great example of this is C-3PO from Starwars, who specializes in 'Human-cyborg relations" and looks like a human. Humanoid robots should be used with humans interacting with them on a personal level (I.E. coitus bots, translater droids, servants, robotic child companions). If the robot is not meant to interact with humans on a personal level, then don't make them look like a human.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .