Robotics: How Do Robots Do Things?

Inside, a robot is mostly the same concept as any other computer-based design:

  • It always requires an input with sensors that detect something that physically changes. This input is often a camera, and often includes sensors along with it.
  • The operating system is relatively similar to any other computer.
  • The processor eventually operates an output that physically does something. It may use a screen, but always uses a mechanical arm, and often has lots of driver “middleware” to accommodate the extra mechanical hardware.

Externally, though, a robot is the ultimate “implementation” of a computer onto the physical world. All other computer aspects are mostly confined to the realm of the mind and ideas, but robots have the means to legitimately do things.

The first industrial robot was Unimate, created in the 1950s by George Devol. It worked on an assembly line at a New Jersey General Motors plant and worked off of drum memory.


Robots are designed for a variety of uses, based on the roboticist’s purposes. However, their uses are always far more limited than someone employed in a relatively similar capacity.

The tasks a robot does best are highly precise, highly repetitive, or require hair-trigger reaction times. One system for controlling them is called computer numerical control (CNC), which specifies with precise mathematical formulas where the computer should go and what trajectory it should travel.

However, robots are terrible at some tasks:

  • The inputs sometimes require further investigation, but it’s not always clear when.
  • The output must be modulated according to an approximate estimation.
  • Absolutely anything involving human interaction, but more on that below.

Because of the excess of details and the very specific use cases of robotic machinery, the most prevalent use of robotics is in tedious factory and warehouse work. In fact, some warehouses are almost entirely automated.

Most robot parts are designed by borrowing directly from nature or existing technology.

Dumb Machines

In many respects, a robot demonstrates how utterly stupid computers are. They are effective at doing exactly what they’re told, repetitively, but don’t understand how to improvise or adapt.

One simple example of this idea is the seemingly easy task of laying bricks:

  1. Lay mortar.
  2. Place brick evenly on mortar.
  3. Lay mortar on the side.
  4. Scrape off excess mortar.
  5. Repeat.

However, after a few trial runs programming a brick-laying robot, you’ll notice it creating bad brickwork. Debugging would show details necessary for a successful task:

  • If any bricks slide from their spot, put them back in their place or adjust other bricks.
  • Press firmly on the mortar to create cohesion, but no so firmly that it offsets the brick from the others too far.
  • If there are any offset bricks, place future bricks at an aesthetically similar height to reflect them.
  • Recognize impurities in the mortar and compensate the brick for it.

All of this would be resolved by the intuition of just about anyone who has been laying bricks for a few weeks.

Computers are the world’s fastest idiots, and robots demonstrate it by understanding specificities without grasping anything in the realm of general principle or intuition.

As another example, imagine instructing someone to pick up pebbles from a flat surface and place them in a jar:

  1. Pick up the pebbles you see over there (pointing finger).
  2. Place the pebbles in that jar there (pointing finger).
  3. When you’re done, let me know.

While small children can grasp this concept easily, effectively programming a robot takes a lot of work:

  1. Demarcate the grabbing zone, indicated as a specific range mapped out from captured camera information.
  2. Set an algorithm that clarifies exactly what a pebble is, then create instructions to grab it, with each grabbing effort modulated with the camera data for every single pebble.
  3. Move the arm to the jar, which must be exactly specified before the maneuver. There should be subroutines for whether the jar is misplaced, full, fell over, or broke.
  4. Repeat 1-3 until all pebbles are in the jar, then clarify a stop sequence for the arm to rest.

Robot AI must also incorporate elements of kinesiology to manage complex tasks.

Less Dumb

One of the easiest workarounds for dumb robots is to have robots operated by people who remotely control them:

  • When a location is too unsafe (e.g., near a bomb, drilling), remote-control robots have been popular for decades.
  • Robotic hands and legs may soon become novel prosthetics technology for disabled people, assuming the limb can accurately read and send nervous system signals.

Lately, machine learning algorithms have dramatically improved the variability of the robot’s instructions. They still have some trouble improvising while crossing difficult terrain, but can balance enough to ride skateboards or zip lines.

But, they’re still relatively useless when completely unsupervised, and are still dumb enough for an errant cardboard box to force them to stop.

The largest impediment to robotics advances, however, is in how absurdly difficult it takes to program seemingly simple aspects that intuition naturally provides:

  • The logic necessary to frame robotic instructions is conventional software design work.
  • The input is very fuzzy, and almost always analog.
  • The output typically must fulfill very precise specifications.
  • It’s difficult to correctly account for all the “edge cases” you can’t know while programming a robot.
  • Even after many iterations, the work to revise a robot’s code is never technically done.


Beyond reality, robots have a tremendous draw in the world of fiction, and people like to imagine rather dramatic (and unrealistic) portrayals of what robots could do:

  • All low-skill jobs replaced by robots, to the point that unions tend to fight their use.
  • Human-looking robots in customer service roles.
  • Use cases in the sex industry.

However, beyond the above concepts about how dumb they are at tasks, there are specific problems for many of the perceived domains:

  • In the event of a poorly programmed robot, high-risk robotic implementations (e.g., cars) won’t withstand human nature’s tendency to blame when people die, so it’s becomes a heavily politicized situation far more than simply upgraded technology.
  • To make a human-like science fiction robot (e.g., Asimov’s “I, Robot”) would require adding the fields of psychology and sociology along with all the other disciplines required for robots, meaning robots would have to first be ubiquitous everywhere to approach that challenge.
  • Widespread adoption will always be suppressed from the same issue as realistic game development: the uncanny valley:
    • Robots are more relatable as they become more human-like, and they’re very relatable when they’re precisely human-like.
    • Right before they’re fully relatable, there’s a odd point where their behaviors and mannerisms are terrifying and creepy.
    • At that point, a robot sabotages all human connection it had gained, and will create a prolonged state of unease to the observer.

In particular, the uncanny valley can turn off the public, since they’re conditioned against it further through the genre of horror stories. Human expressions are vastly complex, with caveats and precise timings for absolutely everything, and robots simply can’t compete with the intuition required to accomplish something that could pass as human.


Like all other technology, robots have a more profound effect than people who discredit them believe, but that profound effect happens over longer periods than the apologists would be quick to assert.

In spite of the sillier trends attributable to robotics, there have been changes to society:

  • Robots have very specific, purposed uses in realms like horticulture, factory work, military support, and navigating small corridors. Many of them are remote-controlled, but they serve as an agent of the human who uses them. This arrangement has been positive.
  • Skills do get replaced, which frees up people to do other things. People find meaning in work, and they do something else that can add value, though older people surpassing about age 40 can be left behind from the change.
  • Lower-intelligence jobs (i.e., for people below an IQ of 85) have almost ceased to exist, since robots can do the job. Nobody has found a workable solution for this.