Month: June 2020

In my Terminator review and my analysis of what a fully-automated tank would look like, I mentioned that human-sized, general-purpose robots that can do the same physical tasks as humans will not necessarily look like humans, or even have humanoid body layouts (i.e. – head, large torso, two arms, two legs). I’d like to explore that idea in greater depth, and to offer educated guesses about what such robots would look like.

First, bear in mind that there are already countless numbers of robots in the world–overwhelmingly in factories and controlled work settings–and almost none of them are humanoid. Instead, their body shapes are entirely dictated by their narrow functions. For example, a robot that welds the seams between two sheets of metal comprising part of a car’s frame will resemble a giant arm and will have a welding torch for a hand. Since it is meant for use in a car factory assembly line where unfinished car frames will be delivered to it via conveyor belt, the robot won’t need to move from that spot, and hence won’t need legs or wheels. And since the act of welding a seam isn’t that complicated, it won’t need a giant computer brain, meaning it won’t have a head. Likewise, a robot designed to move supplies like medicine and linens throughout a hospital will take the form of a large, hollow box with wheels.

Even as robots get cheaper and more advanced in the coming decades and take over more jobs, the vast majority of them will continue looking nothing like humans, and will be designed for specific and not general tasks. Fully-autonomous vehicles, for example, will count as “robots,” but will not resemble humans.

That said, I think “overspecialization” of robot designs will prove inefficient, and that there will be niches for general-purpose robots in many areas of the economy and ordinary life. Some of these general-purpose robots will be about the same sizes as humans, but they won’t look exactly like us. Consider that the humanoid body layout is inherently unstable since it is top-heavy and only has two legs to balance on. If we had millions of bipedal, human-sized robots walking around and intermixing with us in many uncontrolled environments, there would be constant problems with them falling over (or being pushed over) and injuring or killing people. Something like a 250 pound Terminator made of hard metal would be a lawsuit waiting to happen.

Off the bat, it’s clear that general purpose robots can’t be so heavy that, if one fell on you, you would be seriously hurt, and/or unable to push it off of your body. At the same time, it can’t be so light that it tips over when carrying everyday objects like full trashcans, or is even at risk of being toppled by wind gusts. Splitting the difference between the average weights of adult men and woman gives us a figure of 180 lbs, which I think is a good upper limit to how much the robots could weigh.

Also off the bat, it’s clear that the general purpose robots should have the lowest practical centers of gravity and need to have soft exteriors to cushion humans against collisions. A low-hanging fruit helps us solve the first requirement: delete the robot’s head. This might sound very weird, but if we’re unbound by the constraints of biology and are designing a robot from metal and plastic starting from a clean slate, it makes perfect sense.

Since robots won’t eat, drink, or breathe, they won’t need mouths, noses, or any associated anatomical features found in human heads and necks. And since signals from the robot’s sensory organs would travel to its “brain” at the speed of light, there would be no advantage to clustering the eyes, ears, and brain together to reduce lag (thanks to the slowness of human nerve impulses, it takes about 1/10 of a second for an image or sound that has been detected by the eyes or ears to reach the brain), meaning the CPU could be moved into the torso. Doing that would lower the robot’s center of gravity and give the CPU more physical protection than our skulls provide our brains. (Distributing mental functions among several computer cores in different parts of the torso and even limbs would probably be an ideal setup since it would further improve survivability.)

In place of a neck and head, there might be a telescoping, flexible “stalk” or “tentacle” with sensory organs (camera lens, microphone) at its tip. It could extend and shorten, and swivel in any direction. By default, it would probably be facing forward and raised to the same height as a typical human head so it could see the world from the same perspective as we. The top of its torso might only be 4′ 10″ off the ground, but the stalk would rise up another foot. The sci fi space film Saturn 3 had an evil robot named “Hector” that had a crude tentacle like this in place of a head.

The last safety requirement that I mentioned, the need to have soft exteriors to cushion humans against collisions, could be satisfied by making their outer casings from a spongy material like silicone. However, I think it would probably be cheaper and just as effective to give the robots hard outer casings, but have them wear tight-fitting, padded clothes. The general-purpose robots would know how to wash their clothes in standard laundry machines and would periodically do so. Also, if the padding were made of the plastic foam found in life jackets, it would keep the robots from sinking to the bottom if they, say, fell into a swimming pool while cleaning it, or fell off the side of a fishing boat where they were part of the crew.

The need to protect people from accidental injury will also mean that general purpose robots will be made no faster or stronger than average humans. These limitations would be very helpful to us in a “robot uprising” scenario, but they’d be just as beneficial preventing many kinds of small, mundane accidents that could hurt people. For example, if your robot isn’t stronger than you, it can’t accidentally crush your hand by applying too much pressure during a handshake. If it can’t move faster than a jog, it can’t ever build up enough speed and momentum to collide with you with fatal force.

With these safety requirements in mind, it should be clear why the general-purpose “NS-5” robots in the movie I, Robot was unrealistic. There was no reason to give those robots superhuman speed, strength, agility, and explosive movement. Moreover, they all had hard exoskeletons and walked around “nude,” making them collision hazards. (On a side note, I also thought it was unrealistic that a single company–“U.S. Robotics”–would have an apparent monopoly on the humanoid robot market, and that all humans would own the same kind of robot. In reality, there will be many companies making them in the future, and there will be many different robot models and variants that will look different from one another, just as there’s great diversity in how cars look today.) 

Now that I’ve covered the safety issues general-purpose robots will have to be designed to address, let’s move on to exploring the other requirements that will affect how they will look. Since they’ll have to navigate human-built environments like houses and to fit into vehicles designed for us, they will need legs instead of wheels so they can climb steps, arms and hands for opening doors and using tools, and they will need to be skinny and short enough to fit through standard-sized doorways. The requirement for them to be able to sit in chairs and climb over obstacles like low fences and fallen tree trunks will mean the size proportions of their limbs and bodies won’t be able to stray too far from those of humans. They will need fingers that are as thin as ours to type on keyboards and push standard-sized buttons, but they might not have five fingers per hand (it will be interesting to see what the optimal number turns out to be).

It wouldn’t cost much more money to make the joints in the robots’ fingers and everywhere else double-jointed, and they’d gain useful dexterity from such a feature, so I think it would be so. Pivot joints in the arms and legs would also allow for 360 degrees of rotation, further bolstering utility. At first I thought the general purpose robots would have a second set of arms–for a total of six limbs–so they could be more able than humans, but then I realized how wasteful that would be since so few tasks require them. 99% of the time, the second set of arms would uselessly hang down off the robot’s body and be dead weight.

Then again, that 1% of the time when you do need the extra pair of hands to do something could warrant some kind of engineering compromise. The prehensile sensor stalks that stand-in for heads on our general-purpose robots could elongate and grasp onto things, acting like weak third hands (our mouths do the same, and can hold smell, light objects). Instead of, or in addition to that, the legs at the bottom of the robot could terminate in hands instead of feet like ours. Chimpanzees are like this, and many birds also have feet they use for grasping and walking. The setup would make it harder for the robots to run, and maybe less energy-efficient for them to walk, but we’ve already established we don’t want them to be able to run fast, and many of the tasks we’d use these robots for wouldn’t require large amounts of walking anyway (ex – robot butler in your house). Aside from giving them an extra pair of hands for those rare occasions when they need it, having hands as feet would let the robots pick things up from the ground, climb ladders more easily, and maintain better balance on uneven surfaces like roofs.

It almost goes without saying that the robots would be able to walk on all fours about as well as they could walk on two legs. If they weren’t carrying anything and were just going from one place to another, walking on all fours would be safest since that would minimize the risks of them losing balance and crushing someone or breaking something. This is again reminiscent of chimps, and I think the robots might use their “knuckles” when walking on all-fours to keep the palms of their hands clean and undamaged. And interestingly, in laying out this new requirement for optional quadrupedalism, the hypothetical general-purpose robot’s design has superficially converged with the real-life “Spot” robot, made by Boston Dynamics.

One thing I don’t like about Spot’s design is that its torso is a single, rigid piece. The general-purpose robots I’m envisioning–or at least the more advanced variants of it that will be fielded in the more distant future–will need segmented torsos that let them bend and lean a little in all directions. The flexibility of our spines lets us do this, helping us to quickly make small postural adjustments to balance on two feet. The robots might not need anything as elaborate as a human back made of 33 vertebrae, and, as with the number of fingers, it will be interesting to see what the optimal (or sufficient) number of torso segments turns out to be.

Having a flexible torso, four hands, and four, highly flexible limbs that could bend in more ways than we can would also let the general-purpose robots comfortably touch any part of their own bodies, enabling them to self-repair, which would be an invaluable feature. The swiveling sensor stalk plus tiny cameras built into other parts of its body like the hands and torso would also let it see every part of its own body (cameras built into the hands or fingers would also let it reach inside small, tight spaces and clearly see what is inside, letting it guide the appendage, unlike humans who must blindly feel around in such situations). Contrast this with us humans, who have a hard time touching and manipulating some parts of our bodies (like the spot between our shoulderblades) and who can’t see every part of our own bodies because we have only one set of eyes that are in a head with limited rotation.

On that note, having small cameras embedded throughout its body would also eliminate blind spots, which would improve safety since the robots wouldn’t be at risk of running into humans or objects because they were unseen. Whereas human vision is confined to a forward-facing cone, the general purpose robots would see in a 360-degree bubble. The tip of the head stalk might have the biggest and best camera, but losing it wouldn’t blind the robot.

Having “eyes” in the torso and on all four limbs, along with a distribution of its mind and power sources among multiple internal computers and batteries in each place, could enable such a robot to fix itself even if only one limb were operational and everything else were not. Again, this reminds me a bit of something I’ve seen in the animal kingdom, this time among certain insects and spiders. Because they have less-centralized nervous systems than we, their limbs will keep moving after being severed, and, if they are cut in half across the torso, both halves will continue moving and reacting to stimuli.

Additionally, while the robots wouldn’t need to breathe, they should have an ability to suck in, retain, and expel air. This would allow them to duplicate the human abilities to blow out candles or blow dust off of things, and to make our bodies buoyant for floating in water. Of course, the engineering solutions that will let them do this could be totally different from human anatomy’s solutions. A small hole at the tip of one finger could be used to suck in and expel air, and it could be connected to a long tube that would lead to air sacs throughout the robot’s body, perhaps in places not analogous to where lungs are in our bodies.

The robots would also need to be waterproof. This would save them from being expensively damaged or destroyed by something as simple as rain, and would let them periodically clean themselves off with soap and water. Even without sweat glands and shedding skin cells, robots would inevitably get dirty thanks to dust in the air, splatter from kitchen or bathroom chores, or even mold growth. Being able to use a regular shower or a bucket of water and a sponge to clean themselves would be a very important feature, in addition to their ability to clean their clothes.

Another crucial feature would be a built-in power cord that could plug into standard electrical outlets. It might be stored internally in a small, closed compartment, or might take the form of retractable prongs located in one of the hands or feet. I suspect that, rather than get in your way, general-purpose robots will be programmed to run around your house and do chores when you were away at work or school. That would also be safer since it would eliminate any risk of the robots hurting you by accident while they were working. You would come home each day to a clean house and see your robot motionless in its designated corner or closet, plugged into an electrical outlet to recharge.

I’ve already mentioned the robots would need to have cameras and microphones to duplicate the human senses of sight and hearing, but they would also need to duplicate our sense of smell and taste to a degree. Those two senses can provide valuable information about the presence of poisonous gases, smoke, or spoiled food ingredients, and there are situations where a robot would be grossly ill-equipped to respond properly if it lacked them. Our multipurpose robots would thus need air sampling devices and some type of fluid analysis capability. The same technology found in smoke detectors, carbon monoxide detectors, and military poison gas detectors could stand in for a sense of smell. To crudely duplicate our sense of taste, the robot might have something like a litmus strip dispenser and water nozzle built into one of its hands. It could spray water on objects and then touch them with a strip to “taste” them.

The fifth human sense, touch, would need to be duplicated by pressure and temperature sensors distributed throughout the general purpose robot’s body. This feature would be simple to implement.

In conclusion, I predict there will be a future niche for “human-equivalent” robots that are general-purpose, human-sized, and can do all of the physical work tasks that we can do. That said, those robots will look very different from us, as they won’t be bound by the rules of biology or by the genetic path dependence that locks us into our human body layout. I’ve gone into depth describing one type of general-purpose robot, which could be described as a “headless humanoid.” However, I think robots with other types of body layouts could also fill the niche, perhaps including “centaurs”, “big ants”, and “dogs with one arm on their backs.” Just as there are many types of vehicles on the roads today that fulfill the same roles, I am sure there will be many types of general-purpose robots. I simply don’t have the time to envision and describe what each one could be like.

General-purpose, human-sized robots will of course not be the only kinds of robots we’ll mix with on a daily basis in the future, and in fact, I think they will be outnumbered by other, specific-purpose robots whose forms reflect their specialized functions. Self-driving cars and autonomous lawnmowers are good examples.

Finally, the general-purpose, human-sized robots must not be confused with androids, which will look identical to humans. I think the general-purpose robots will be used for jobs that don’t require anything more than superficial interaction with humans, like scrubbing toilets, restocking store shelves, and fixing appliances. Androids would be built to provide companionship, and to do service-sector jobs where warm and personable service was expected. If your beautiful android spouse broke, then your grubby, headless, weird-looking robot servant would fix it.