Monday 7 April 2014

What's in a word?

Languages are funny creatures. They defy control, evolve into the strangest of things, and refuse to obey anything but the simplest of rules as to how words become adopted and adapted.

Recently I've been thinking about robots.

Strike that. I've been thinking about robots for a long time, and more recently my Martian-in-crime wondered as to what connotations are associated with the names that we have for robots, and how does the name of a thing alter our perception of it? I still think that the cognition depends on some strong way on the ability to categorize and perceive items and concepts. So in some sense the naming of a thing is a critical step - as names, even invented ones, will always carry some resonances with existing words in a language. I may not go so far as to take this to the end point of Le-Guinism, where the truename is a sort of undisclosed API that allows reality to be subverted. But it seems obvious that the act of naming a character Voldemort is not a naïve choice to anyone with a passing knowledge of indo-european languages.

Thus, prompted by my Martian, I wondered about the systems that we're building that have lifelike characteristics. One might be tempted to broadbrush them all as being 'robots', but let's unpack that word. You may know that this is a fairly recent neologism, coined by a Czech chap in the '20s. What you may not know is that it wasn't Karel Čapek, but his brother, Josef Čapek, who invented it, drawing on the perfectly good pan-Slavic word robota that means drudgery or servitude.

So we have this word: robot. And along with it come connotations of serfdom and slavery. If we label complex systems as robotic then we also constrain our expectations of them and will perpetuate the master/slave relationship. Such a dialogue is probably harmless when the worst that your house can do is to burn your toast, but subconsciously treating your Doctor as a mechanical peon probably will do you no good.

So how else might we describe a synthetic complex system?

Well, we might describe it as being:
automatic from αὐτόματον, (pron. automaton), which means "acting of one’s own will"
or even
mechanicalfrom μηχανικός (pron. mekahnikos), which means "a device or tool"

But mechanical carries with it a sense of rigidity and inflexbility - which is probably a good thing sometimes. You don't want an elevator to have emotions.

Even automatic, which at least has some semblance of self-direction, has mutated in English to be almost a perjorative term.
"I was on auto-pilot M'lord, I didn't know what I was doing!"
And so it goes.

I suggest that English lacks a word to describe the autonomy and complexity of probable artificial life-like systems. To describe a postulated 'Google car' as being automatic is to overlook its ability to react like a person would to changes in a road's state. To describe an embodied synthetic intelligence as being an 'android' is to suggest that it ought to be manlike in some way, and if the form that it operates is not shaped like a human, do we have to use unwieldy concepts such as an android-octopus?

I further suggest that we need a new word to cope with the probable rise of synthetic embodied intelligences: note, I deliberately avoid the use of artificial as a description. We don't talk about artificial flight when describing the motion of an airplane when compared to that of a bird. Both fly, even if the bird does so in a different manner than that of an airplane. But only one is a synthetic actor. Similarly, a synthetic intelligence may appear to perform thoughtful acts, despite it doing so without goo and meat.

So - I open the floor. What would you call a system that may or may not be embodied, may or may not take the form of a human, and yet appears to have free-will and perform thoughtful activites?

Answers on a post-card please.