What makes a machine a robot?

You often hear about companies using “robots” in their factories. In the factory where I work, all of the machinery has aspects that are controlled by computers, such as speed and ingredient mixes, but nobody ever refers to them as robots. Does a robot have to have AI? Or is it just a matter of semantics?

In my humble opinion, based on 60 years reading science fiction, a robot has to sense it’s surroundings and make decisions based on that input to be a robot. Bomb Robots controlled by a person viewing a monitor and controlling via a joystick are not robots to me.

Robot is not a well-defined term. Give something a mechanical arm and it’s likely to be called a robot, even if it’s completely remotely operated or preprogrammed.

Agreed. Those should be called drones.

As you are asking about the meaning of a word, the question you raise – whatever its answer may be - is, indeed, a matter of semantics.

I work in manufacturing too. We don’t call our machines robots, we call them tools. However, there are certain parts of the tool that are called a robot arm.

Electronically programmable, unless it is some sort of mechanical or optical (future stuff, wait and see) programmable ones.

40 years ago they were speaking of robot radio stations. I assumed at the time that they were stations that were completely operated from a remote location, but don’t know.

As stated above, the term is not well defined, but if it has a multiaxis, programmable “arm” the term “robot” is usually used. 25 years ago where I worked, the term “flexible automation” was also used interchangeably with robotics. When I get an automated phone call, I say that a robot called me.

Yes, robots are automatons–working ‘independently’ of human controllers, but according to fixed instructions.

Doesn’t it boil down to having sensors and making decisions?

I seem to remember early spacecraft that crashed into the moon, etc were often called “robots” or “robotic control”.

The things that make automated, pre-recorded-voice phone calls are called robots. So are the “web crawling spiders” that compile the data for search engines (and, of course, despite the talk of “crawling the web” they are really just programs that sit on one computer at Google or wherever’s HQ, and check out lots of web pages and tehe links they contain).

Some people might say that to be a “real” robot it must be autonomous, and have AI, but, unfortunately, it is probably at least as hard to pin down what counts as AI as what counts as a robot.

A fixed program is not a robot. An autonomous (during its work) physical unit, running a fixed program, is a robot. It may have ‘senses’–it must receive some kind of input to respond to–but all its possible responses are ordained.

Nor is a fixed program (embodied in a robot or not) AI. Intelligence means being able to handle circumstances not accounted for in ‘programming’–to learn, to adapt.

The definition of “AI” is far more slippery than the definition of “robot”. In practical terms, the definition most often used for “AI” seems to be “what we don’t have yet”. Back before the first chess-playing programs, folks said that chess would require AI. Then, once we passed that milestone, they said well, that’s not really AI. Nowadays, we have devices that you can talk to and ask for the weather next week in Bangkok or the outcome of the 1972 World Series or the integral of z^2 dz whatever other piece of random information you want, and people still don’t consider it AI.

Nor do you, right? You know (approximately, anyway) how that clockwork spins.

The closest generalization that seems to explain what people call robots is that it is a machine that does a job a human used to do. That covers everything from mindless repetition on assembly lines to nursing units.

On the other hand, the subroutines that control the behaviors of NPCs in computer games are conventionally called AI.

Mind you, given that Simon, Shaw and Newell wrote a General Problem Solver program back in 1959, one might wonder why we still have any unsolved problems.

The meaning of AI is slippery, because it has been a deliberately vague, hype term from the beginning.

One of the things that has come out of AI research is the realization that we don’t actually know how natural intelligence works and while some are still working on the “mimic a brain” idea, most of today’s “artificial intelligence” - like Google’s speech recognition - achieve their results in ways that don’t mimic human thought and so people are loath to call it “intelligence” of any sort.

Personally, I think computer intelligence will always be something that achieves human intelligence like results in a completely non-human like way.

And this why the British press talked about a robot invasion in the 1940s! Because those robots did follow simple mechanical programs and were not controlled remotely:

From the QI tv quiz show: