The Life Electronic: Turing results and “Her.”
Today's rambling comes after watching the movie “Her” (2014) and in light of recent news that an application has “passed” the Turing Test.
The latter first: the ephemeral significance and permanent unimportance of the Turing Test and any current attempts has just smacked me in the head. The fact that this latest “success” did so by credibly imitating a 12 yr old Ukrainian boy makes it a perfect example of the mechanical Turk. The Turing Test will, for a considerable time into the future, remain like the “strapped chicken” test for the Star Wars missile defense system: a demo so far removed from the reality it claims to prove that only the most casual observer (aka “tax payer” “voter” “TV watcher”) would think it worked.
In the movie, the number normal human things the software does is ridiculous. The plot device (must I say “spoiler alert”) is silly because the OS already performs a million tasks between words uttered. Scaling it up to a billion—or even “times a billion”—would not materially change the behavioral adaptation the OS would need to work in the first place. It is not that she goes to a place between the uttered words, it's only that the author finally got there in Act 5.
And while I'm going on about 2 creatures of vastly different life-spans and -scales interacting, let's talk about emotions. They're the high-speed calculator for our brains; a simplified system that takes input and rapidly guesstimates an outcome, and signals the higher functioning system with this limited vocabulary. Happy. Sad. Scared. Mad. Our emotional system is this ancient part we inherited from the reptiles. So the OS having one is lame. Building it to simulate emotions for interactions is also silly. Data would no more have an emotion chip than need one. If you want your AI to have a limbic response, you could build one, but it would be like putting chrome bumpers on a 747. Where would you put them, to what end, and what were you expecting.
My point is this: we're a long way from knowing what our first AI will even look like, but it sure won't look like us.