-->

Saturday, August 17, 2013

Are Humans In Jeopardy?

Is the era of human supremacy coming to an end?

What to do now that computers can beat us at our own game. That game, of course, is "trivia," the ability to access random information on a variety of subjects, which would seem to be a prototypically human strength. Yet the contest on TV's "Jeopardy" was one of the most foregone conclusions ever presented for our viewing pleasure.

Following months of hype, Watson - IBM's supercomputer - faced off against the same two contestants it had already trounced in a trial run. After all the publicity, watching the event itself was kind of like seeing the Packers and Steelers on videotape a week after the Superbowl. But this latest victory is still a milepost in the history of artificial intelligence.

Artificial Intelligence and Chess

It's a victory that has been predicted since the earliest computers were in their infancy. Alan Turing, whose pioneering work with logic and theory laid the foundations for computer design (and helped to break the German Enigma code in World War II), could see no reason why his electronic offspring should not eventually meet and even surpass the capabilities of those who designed them.

That was quite a stretch of the imagination back when computers were little more than rapid calculating machines. Others at the time (the 1950's) believed there were certain types of thinking in which humans would always excel. Oddly enough, chess was often cited as an example. But as we now know, the game of chess is trivial for computers, given enough memory and processing power.

By the 1980's futurist Ray Kurzweil predicted that a computer would be able to defeat the best human chess master by 1998. He was wrong, because it happened in 1997 when IBM's Deep Blue overcame Russian grand master Gary Kasparov. By this time most of us were not even surprised, because in most cases we ourselves could not defeat a toy sold by Radio Shack.

If humans still had any claim to unique capabilities, it may have been in our ability to rapidly synthesize a solution based on hazy and tenuous relationships between apparently unrelated things. Though typified by parlor games like "Trivial Pursuits," the same mental ability gives us creativity, the capacity to invent, write music and literature, to produce art and original scientific theories.

But now we may have to cede this advantage to the next generation of supercomputer. Recently Kurzweil picked Watson as the favorite in the Jeopardy competition, saying that even if it lost it would soon improve enough to come back and win. At the rate they're going, he is convinced that within a few decades they will exceed the memory capacity and thinking ability not just of a single human, but of all humans combined.

Passing the Turing Test

What's next for artifical intelligence? Maybe all that's left is the so called "Turing Test." Alan Turing proposed setting up a pair of teletype machines, one leading to a computer, the other to a human operator. If someone conducting a conversation by teletype could not tell the difference between them, then the computer would pass the test - meaning that it had demonstrated human intelligence.

That sounded far-fetched at the time, but those working in the field of AI have come to believe the Turing Test is actually too easy. In fact, demonstrations have already been done which were good enough to fool a certain percentage of human subjects. But the methods used were not nearly advanced enough to be considered true thought.

So other tests will no doubt be devised, with computers - or whatever they come to be called - rising to the occasion, and passing over one bar after another. IBM's room full of computers will shrink to the size of a chip, as it has done before. And perhaps one day computers will be the ones proposing the tests, asking us to demonstrate what it means to be "human."