The Turing Test

Will computers ever become more human?  How smart are computers, anyway?

I don’t know about yours, but mine computer’s pretty dumb.  I mean, if I want to count something, like the balance in my checkbook or the scores in my grade book, my computer’s great.  Better than me, that’s for sure.  I remember what one of my abstract algebra professors used to tell me: there are three kinds of people in this world: those who can count and those who can’t.  I know which type I am, and it’s a darned good thing I’ve got a computer.

But a computer isn’t smart.  I’d never mistake my computer for, say, my partner.  My computer never forgets anything, for example.  Thank God Gene forgets, else he’d never be able to stand living with me.  But Gene is different from my computer in other more lovable ways too numerous to mention.  I’d never mistake him for my netbook.

Still, some futurists think that one day computers will assume human-like qualities.  Apple’s Suri tries to simulate a human being.  I’ll admit, she seems to be as hearing impaired and stubborn as I’ve gotten as I age.  Still, no one would mistake Suri for a human.

So, how will we know when computers have passed the boundary into consciousness and joined humanity?

Alan Turing thought deeply about questions like this, in between solving hard mathematics problems like Hilbert’s Entscheidungsproblem and deciphering the Nazi Enigma code during World War II.  He wondered if machines would ever be able to think. I wonder sometimes if people can think, but that’s another question.

Turing decided to rephrase the question to one of imitation: is it possible for computers to imitate a conversation so well that a judge could not distinguish machine from human? This is the Turing Test.  The original idea was for a text-based conversation, so the judges base their assessments only on the words on the page (this was 1950, before CRTs).

This has nothing to do with the correctness of the answers the computer might give.  Winning chess has nothing to do with the Turing Test.  The issue is whether or not the machine’s end of the conversation seems human to an impartial judge who doesn’t know if the conversation is with a machine or a person.  If the judge can’t tell the difference, then the machine passes the Turing Test.

The test has well-documented limitations, and it’s not entirely clear what it’s really measuring.  However, it’s generated much discussion about what it might mean to be conscious or human and whether or when machines might be able to “pass” the test.  Two famous technologists even have a bet on whether or not a machine will pass the test by 2029.

So this morning I stumbled across this article.  It seems that researchers have programmed a computer to simulate a person playing a computer game. I suppose a curmudgeon might argue about whether that constitutes human behavior, but for the sake of argument let’s suppose it does.  These researchers report that impartial judges find that the machines exhibit human gaming behavior 52% of the time.  That’s apparently better than real human gamers, who exhibit such behavior only 40% of the time.

Apparently these judges have met some of the people my daughter dated in high school.  Come to think of it, those boys were more or less like me in high school, so strike that.  They were consummate humans.

So does that mean that the above bet is settled and we’ve got a computer that passes the Turing Test?  Well, no.  These judges were looking at behaviors of avatars in computer games, not at the give-and-take of a conversation. It’s hard to imagine the conversational brilliance of, say, Dorothy Parker popping up in a computer game.

Thinking computers, complete with personalities, are fun to imagine.  They have become common in science fiction, from Asimov in I Robot, to Hal in 2001, to the android David in Prometheus.  We like to anthropomorphize our machines, like Stephen King in Christine.  We imbue them with human qualities to make our world more bearable.


A machine that is truly human opens huge potential, both for the machine and for us.  Keith Laumer’s Bolo series gave us a glimpse of this as early as the 1960s, and today visionaries are pushing the technology boundary beyond our imagining.

Do I think machines will become human?  I think the answwer is yes, but it’s impossible to say when.

Right now, I’d just settle people being more human to each other.

 

 

This entry was posted in Uncategorized. Bookmark the permalink.

Leave a Reply