I once had a robotics professor who argued any definition of 'robot' broad enough to include roombas and self-driving cars and industrial robot arms (even if made narrow enough to exclude remote-control toy cars, cars with cruise control, CNC mills and washing machines) would inevitably define cruise missiles as robots. Meaning the world had 'killer robots' before we had digital computers.
Instead, he argued the definition of a robot is "I know it when I see it"
> Also I'd argue that a CNC machine is a robot as well, no different than a programmable industrial robotic arm.
It's true - a robot arm is not so different from a CNC machine.
But a 5-axis CNC is not so different from a 3-axis CNC, a 3-axis CNC is not so different from a laser cutter, a laser cutter is not so different from a vinyl cutter, a vinyl cutter is not so different from a large-format printer, and a large-format printer is no so different from an inkjet printer.
In general the game of necessary and sufficient conditions for any non mathematical concept will always lead to inconsistencies, there's no such prescriptive "definition" for most words. try defining "game", or "chair" in such a way!
When I was a kid my dad had an HP-7475A pen plotter. When explaining this device to friends today, I describe it as "a robot that draws with a pen". It's got a little robot hand and everything that can pick up and replace one of six pens.
> neither versatile in its objective, nor programmable.
how programmable does it have to be for the robotic classification to stick? A cruise missile also cannot be programmed to be general - you can only program it to follow a trajectory. Just like you can only program a dishwasher to do dishes in a particular way.
Autonetics designed the inertial guidance system for the Minuteman 1 ICBM. They also sold it as a general purpose computer and distributed to Universities to learn about general purpose computers.
It took realtime input from gyroscopes and starmaps and calculated output for the thrusters.
No? You’re thinking of a rocket-motored missile. Cruise missiles are little planes. They can often autonomously navigate terrain. A cruise missile is more a robot than a dishwasher, even a smart one, particularly modern ones which do target identification onboard.
> classifying cruise missiles as robots, but failing to make the same judgement on a dishwasher is biased thinking
'Cruise missile' and 'diswasher' are about as broad categories as robot versus not robot. (If you're calling out 'biased thinking,' say why it is. Otherwise it sound slick you don't know what bias means.)
> about as broad categories as robot versus not robot.
there's no way cruise missiles and dishwashers as categories are similar in broadness to the categories of "robot" vs "not robot".
The point i was trying to make is an attempt to tease out the bias in the original post i am replying to - that of having the robot moniker assigned to things they fear more, instead of a more objective classification.
> there's no way cruise missiles and dishwashers as categories are similar in broadness to the categories of "robot" vs "not robot"
…why?
> that of having the robot moniker assigned to things they fear more, instead of a more objective classification
Sure. I’m rejecting that oversimplification.
Solving a Sinner circle is fundamentally simpler than even a basic flight-control system. The examples you chose refuse your comparison. That suggests your model is wrong.
Most robots are programmed to do one thing well. You could program a cruise missile to fly around in circles or chase the sun, just like you could program an industrial robot arm to wave like a music conductor, but they don't because that doesn't match the goals of their users.
I think the distinction is about the interaction with physical world. I robot must change its own position and/or can be able to manipulate other objects.
A washing machine does’t care about the pose of the laundry inside. No direct manipulation. It just rotates the drum with a motor.
A self driving car locates and moves itself. A robotic arm manipulates objects. They are like half-robots. A “robot” robot can do both.
My washing machine detects when the wet laundry inside is imbalanced and will do a slow half rotation to evenly spaced the wet laundry before a spin cycle
But actually my washer does respond to to load levels that it senses, then fills with different amounts of water accordingly.
And then like an unruly robot, it sometimes becomes unbalanced and "ambles" across the floor unexpectedly when it's supposed to be a completely stationary automaton.
A cruise missile being a robot is not far-fetched, I’d say. In Swedish both cruise and ballistic missiles (and other similar munitions) are called robots within the military.
Roombas, self driving cars, industrial robot arms (depending on what they do), cruise control,cruise missiles, are all robots by the dictionary definition. They react to their environment. That's the classifier.
His view of I know it when I see it is nuts but he is right on many of the classifications. Some were incorrect though. But a robot is just anything that has a sensor to interact with the environment.
We have simply changed the meaning to be more human like which is not accurate.
> They react to their environment. That's the classifier.
An industrial robot arm, out of the box, only has internal sensing (joint position, torque etc) and a start/stop button - can't sense the external environment at all.
But I've got a light on my porch that turns on when it gets dark.
By the 'react to the environment' definition the former is not a robot, and the latter is.
The distinction between internal and external variables is interesting in its own right, but both devices are technically robots. A thermostat is also a robot.
We're starting to need gradual words to linguistically orient ourselves: automaton, drone, robot, android, etc. Where "robot" is a narrowing of the technical "feedback loop device".
It would have to have a force torque sensor which interacts with the environment around it (aka the object being manipulated). Hence it being a robot arm.
However if indeed an industrial arm doesn't have that, then yes robot is the wrong term for it.
"I know it when I see it" has to be up there in terms of useless thought-terminating clichés together with "no smoke without fire" and "if he wanted to he would"
I think the modern day equivalent is the constant huffing and puffing around where to draw the line for tech legislation, with the implicit message that 'if a line is hard to draw, we should not draw it' and a constant bad-faith battle for perfectly defined legislation.
The answer to 'where do we draw the line for this thing that has widespread negative effects' is always "somewhere".
It goes a bit further than that. If there isn't a somewhat objective definition then you can't ever be sure that two people are talking about the same thing. Sharing the same context isn't enough to reliably get a convergence of beliefs about whether something falls into a category. And it is a mistake to start legislating about it because the laws are going to be meaningless.
What I mean is that throwing up your hands and saying 'I know it when I see it' is entirely pointless. There's a reason scientific papers have an (often long) section for definitions.
It's acknowledging that this is a place where rigid, universal definitions, that include everything that is X and exclude everything that is not X, are not possible. That human judgement is necessary to determine what is or isn't X.
This doesn't mean that we have no guidelines whatsoever for how to recognize X. We can list qualities that a robot either must have, or should probably have, and anything that definitely doesn't have them probably isn't a robot, and anything that has enough of them probably is a robot.
I just don't think we'll be able to agree (and that's okay). I'll just put my thoughts here, not expecting that you'll meet me anywhere.
Human judgement in itself is not uniform or defined enough for it to be useful, but dispersed and more random than just situational. And yes, universal definitions in general are extremely rare outside of a subset of definitions in hard sciences. The only reasonable thing we can do is agree on a definition in a certain context, ie definitions in a scientific paper or in a conversation. But 'I know it when I see it' or some vague idea of human judgement doesn't even work as a definition within a defined context, and therefore doesn't help us find common ground or make a conversation useful. Which is why I characterize it as a thought-terminating cliché.
In a way, this entire back-and-forth is an example of us 'knowing what a thought-terminating cliché is when we see it' which unsurprisingly, to me, is unhelpful.
Common speech is slippery, definitions change over time, and the only way we can define words is to see how people use them. In technical context we do define words so that everybody knows what we are talking about.
A good example is the use of the term AI, and how we use it in CS vs how the general public understand it.
I think there's a meaningful difference there: "thought-terminating cliché" specifically means something that causes those who say it (or accept it) to stop thinking there. That's not just something that's up to interpretation: that's something that has actual variation depending on the people involved.
To put it another way, if we posit Person A and Person B, they might both describe "I'll know it when I see it" as a thought-terminating cliché, because they've heard it described as such, and Person A will actually leave it at that, not seeing any reason to discuss the subject further. But Person B does want to continue discussing, and certainly continues thinking about it. Thus, even though they saw the phrase as a thought-terminating cliché, for them, it actually wasn't one.
On the other hand, terms like "robot" and "pornography" really are purely something that's up to interpretation. If I say "I think a light with a motion detector is not a robot", there's no more objective meaning of "robot" that anyone can point to to definitively say I'm wrong.
A Supreme Court judge being reluctant to define a social norm makes sense to me. Other future judges have the ability to update their perception. Prescribing social norms to a textual definition is a problem as social norms change.
Robot etymology is from Czech robota, robotník, robotiti. To work, worker, to do work. I especially like "road works" signs in Poland saying "uwaga roboty drogowe" and translate/visualise it in my mind like "Danger robots on the road"
On a light tangent, aren’t Eastern Europeans bothered by the fact that the word “slaves” was literally coined after them? Is there any discourse locally around any of that?
A food processor and stand mixer are robots in French[0]. I suppose this rather supports the claim that "robot" is a social construct as different societies have constructed it differently. N.b. I haven't taken a French class in almost 30 years. Please excuse any nuances in the French use of "robot" in this context that I may have missed.
ETA: I almost forgot! Our nearby hardware store has a robot that sharpens kitchen knives. It's much less impressive than that sounds, both in appearance and how well it actually sharpens a knife. Nevertheless, I did note to the person operating it that if the robot uprising comes, we may regret the choice to simply give the robots knives.
I strongly advise the podcast "Robot Or Not". The hosts have very strong opinions on what exactly constitutes a robot, and readers write in with edge cases. It's been going on for a long time, so nowadays they also categorize other things.
One time I was messing around with the settings for a 3D printer in whatever software I was using to control it, and found a dropdown for "Machine type" or similar that defined my printer as a "3-axis robot." That gave me pause. Is it a robot? Maybe. That printer didn't even have any sensors, apart from the thermistors that fed into the temperature control PID loops. But by someone's definition, it's a robot.
That was and remains fun to think about. I guess that like the author, I will "be pedantic all day long."
There's an old saying, "if it works it's not AI". We could make a similar statement: "if it works, it's not robotics". That is, once a robot becomes reliable and efficient enough for everyday use, we no longer consider it a robot we just consider it a machine or tool.
Dropping the “universal” was a mistake. Even a robotic arm that just performs one task isn’t really a robot, it’s just a more capable CNC machine in a different form factor.
Sorry to be pedantic - but the correct naming is "kamikaze drone" or "suicide drone" when The Bad Guys (tm) use them and "Cruise Missile" or "Precision Drone" when The Good Guys (tm) deploy them.
the common distinction between robot and machine is that machines are mostly stationary and don't move by themselves. (Self-driving cars sit in between, depending on further developments, they might slide into the bot category.)
Indeed. "Robot" is a radial category that includes machines that resemble humans in either appearance or the sophistication of tasks they can perform. It's one if those terms like "fruit" that doesn't have a strict technical definition. A robotic arm is described as robotic because of its resemblance to a human arm. If you put human-like limbs on, say, a tank (as in Metal Gear, or Tomy's Battle Skipper line of toys from the 90s), the tank becomes more "robotic" as a result. A self-driving truck is considered a robot because its operation incorporates planning, navigation, following of rules including traffic laws, and responding to stimuli such as traffic and people/animals in the road, all undertaken autonomously and resembling the kinds of decisions a human must make when driving a vehicle.
Oh! TFA is written by Mikell Taylor. I worked with her once, yonks ago.
Roomba and the likes are definitely robots. The advanced ones have cameras, map your floor, vaccuum clean, water clean, etc. You can contact them remotely (a friend uses her cleaning robot to spy on her husband while he's at home!) and give them orders.
Nobody ever said "robot" meant "machines with AGI thinking": robot, literally as defined by the person mentioned in TFA and who coined the term, meant "to work" / "forced labor". Work/labor that could be done by servants.
If cleaning my entire apartment by vaccuum cleaning, water cleaning, giving me reports about it, obeying my orders (e.g. if I tell to stop) ain't "automated labor that could have been done by a servant", I don't know what is.
Instead, he argued the definition of a robot is "I know it when I see it"
Also I'd argue that a CNC machine is a robot as well, no different than a programmable industrial robotic arm.
IMHO, a dishwasher isn't a robot because it is neither versatile in its objective, nor programmable.
It's true - a robot arm is not so different from a CNC machine.
But a 5-axis CNC is not so different from a 3-axis CNC, a 3-axis CNC is not so different from a laser cutter, a laser cutter is not so different from a vinyl cutter, a vinyl cutter is not so different from a large-format printer, and a large-format printer is no so different from an inkjet printer.
In general the game of necessary and sufficient conditions for any non mathematical concept will always lead to inconsistencies, there's no such prescriptive "definition" for most words. try defining "game", or "chair" in such a way!
how programmable does it have to be for the robotic classification to stick? A cruise missile also cannot be programmed to be general - you can only program it to follow a trajectory. Just like you can only program a dishwasher to do dishes in a particular way.
It took realtime input from gyroscopes and starmaps and calculated output for the thrusters.
https://en.wikipedia.org/wiki/D-17B
No? You’re thinking of a rocket-motored missile. Cruise missiles are little planes. They can often autonomously navigate terrain. A cruise missile is more a robot than a dishwasher, even a smart one, particularly modern ones which do target identification onboard.
Therefore, i am just arguing that classifying cruise missiles as robots, but failing to make the same judgement on a dishwasher is biased thinking.
'Cruise missile' and 'diswasher' are about as broad categories as robot versus not robot. (If you're calling out 'biased thinking,' say why it is. Otherwise it sound slick you don't know what bias means.)
there's no way cruise missiles and dishwashers as categories are similar in broadness to the categories of "robot" vs "not robot".
The point i was trying to make is an attempt to tease out the bias in the original post i am replying to - that of having the robot moniker assigned to things they fear more, instead of a more objective classification.
…why?
> that of having the robot moniker assigned to things they fear more, instead of a more objective classification
Sure. I’m rejecting that oversimplification.
Solving a Sinner circle is fundamentally simpler than even a basic flight-control system. The examples you chose refuse your comparison. That suggests your model is wrong.
They absolutely do for research and target practice.
A washing machine does’t care about the pose of the laundry inside. No direct manipulation. It just rotates the drum with a motor.
A self driving car locates and moves itself. A robotic arm manipulates objects. They are like half-robots. A “robot” robot can do both.
But actually my washer does respond to to load levels that it senses, then fills with different amounts of water accordingly.
And then like an unruly robot, it sometimes becomes unbalanced and "ambles" across the floor unexpectedly when it's supposed to be a completely stationary automaton.
You can program a diswhasher to brew beer.
Oh, Jesus. Here we go. I have no doubt that it has been done.
Googles a bit...
Could not find anything other than people using dishwashers to disinfect beer bottles.
You can't leave us hanging like that, dude. Please tell me it is true.
https://www.techspot.com/news/4230-automated-beer-brewing-wi...
Who ever controls the CNC machines will win The War Against the Machines.
His view of I know it when I see it is nuts but he is right on many of the classifications. Some were incorrect though. But a robot is just anything that has a sensor to interact with the environment.
We have simply changed the meaning to be more human like which is not accurate.
An industrial robot arm, out of the box, only has internal sensing (joint position, torque etc) and a start/stop button - can't sense the external environment at all.
But I've got a light on my porch that turns on when it gets dark.
By the 'react to the environment' definition the former is not a robot, and the latter is.
We're starting to need gradual words to linguistically orient ourselves: automaton, drone, robot, android, etc. Where "robot" is a narrowing of the technical "feedback loop device".
However if indeed an industrial arm doesn't have that, then yes robot is the wrong term for it.
I think the modern day equivalent is the constant huffing and puffing around where to draw the line for tech legislation, with the implicit message that 'if a line is hard to draw, we should not draw it' and a constant bad-faith battle for perfectly defined legislation.
The answer to 'where do we draw the line for this thing that has widespread negative effects' is always "somewhere".
Something that might have been considered pornography 100 years ago might be considered tame today.
It's acknowledging that this is a place where rigid, universal definitions, that include everything that is X and exclude everything that is not X, are not possible. That human judgement is necessary to determine what is or isn't X.
This doesn't mean that we have no guidelines whatsoever for how to recognize X. We can list qualities that a robot either must have, or should probably have, and anything that definitely doesn't have them probably isn't a robot, and anything that has enough of them probably is a robot.
Human judgement in itself is not uniform or defined enough for it to be useful, but dispersed and more random than just situational. And yes, universal definitions in general are extremely rare outside of a subset of definitions in hard sciences. The only reasonable thing we can do is agree on a definition in a certain context, ie definitions in a scientific paper or in a conversation. But 'I know it when I see it' or some vague idea of human judgement doesn't even work as a definition within a defined context, and therefore doesn't help us find common ground or make a conversation useful. Which is why I characterize it as a thought-terminating cliché.
In a way, this entire back-and-forth is an example of us 'knowing what a thought-terminating cliché is when we see it' which unsurprisingly, to me, is unhelpful.
A good example is the use of the term AI, and how we use it in CS vs how the general public understand it.
To put it another way, if we posit Person A and Person B, they might both describe "I'll know it when I see it" as a thought-terminating cliché, because they've heard it described as such, and Person A will actually leave it at that, not seeing any reason to discuss the subject further. But Person B does want to continue discussing, and certainly continues thinking about it. Thus, even though they saw the phrase as a thought-terminating cliché, for them, it actually wasn't one.
On the other hand, terms like "robot" and "pornography" really are purely something that's up to interpretation. If I say "I think a light with a motion detector is not a robot", there's no more objective meaning of "robot" that anyone can point to to definitively say I'm wrong.
It is a non-definition.
ETA: I almost forgot! Our nearby hardware store has a robot that sharpens kitchen knives. It's much less impressive than that sounds, both in appearance and how well it actually sharpens a knife. Nevertheless, I did note to the person operating it that if the robot uprising comes, we may regret the choice to simply give the robots knives.
[0] https://www.boulanger.com/c/robot
https://podcasts.apple.com/us/podcast/robot-or-not/id9979302...
https://open.spotify.com/show/4M2AELzmQy77FxaYZMl6e9
https://pocketcasts.com/podcast/robot-or-not/b911efc0-e2f6-0...
As are all words.
That was and remains fun to think about. I guess that like the author, I will "be pedantic all day long."
They are all killer robots
https://www.google.com/search?client=firefox-b-d&q=futurama+...
Oh! TFA is written by Mikell Taylor. I worked with her once, yonks ago.
Nobody ever said "robot" meant "machines with AGI thinking": robot, literally as defined by the person mentioned in TFA and who coined the term, meant "to work" / "forced labor". Work/labor that could be done by servants.
If cleaning my entire apartment by vaccuum cleaning, water cleaning, giving me reports about it, obeying my orders (e.g. if I tell to stop) ain't "automated labor that could have been done by a servant", I don't know what is.