Wednesday, February 01, 2006

The 29th Day in the Lily Pond of Artificial Intelligence

Phenomena that grow geometrically remind me of the 29th Day Riddle. As the speed, power and memory capacity of computers contribute to the rise of artificial intelligence, the potential for geometric advances calls to mind the shock value of the 29th Day. I originally learned the riddle in an Ecology class in 1974. It goes: A lily grew in a pond. Every day each lily produces an offspring, thereby doubling the population. If the pond fills with lily pads in 30 days, when was the pond half-full? The answer is, surprisingly, the 29th day. Such is the nature of processes whose products grow geometrically.

Ecologists use the riddle to make a point about resource use and environmental disaster. The question is, when the pond was half-full after 29 days, did the pond creatures know they were on the brink of total lily coverage? After all, in 29 days only half the pond’s space had been used. The remaining area equals all that was used in the history of the pond. What’s the problem?

The 29th Day Riddle may inform our thinking about artificial intelligence (AI). The time is coming where computers may replicate or mimic human intelligence. Alan Turning devised the Turing Test to determine when machine intelligence matches the human kind. The first computer to pass it will mark the dawn of the 30th day. Are there clues that we are now at the 29th?

Most AI researchers admit that we do not yet have a full grasp of what human intelligence is. However, what we do know allows us to compare both systems on some basic levels. For example, speed is a factor in intelligent behavior. The person who solves a problem in a minute shows more intelligence than one who solves the same problem in an hour. In terms of raw speed, computers can already execute more commands per second than the biological brains of humans. This already gives computers the appearance of being smarter than humans in very limited ways. No human stands a chance of beating Mapquest in being given two random addresses and finding the way to get from one to the other. But Mapquest is not yet truly intelligent. It just has far more speed and power.

Memory is another factor that more or less indicates intelligence. With the Internet, any linked computer has access to information stored in countless memory systems. When humans die, much of what they know is lost. When machines wear out, their information memory can be simply transferred. Machine “memory” is collective and important information will virtually never be forgotten.

Vast memory combined with processing speed again allows machines to execute seemingly intelligent actions in limited domains. While my computer may not understand English, Alta Vista’s Babel Fish can translate English words into twelve languages faster that I can turn the pages of a dictionary. Google, Netscape and other search engines use the speed of computers to page through millions of web sites in seconds.

Ray Kurzweil in The Age of Spiritual Machines: When Computers Exceed Human Intelligence, predicts that the 30th day may dawn about 2015. According to Moore’s Law computer power, like lilies, doubles every “generation.” Unlike lilies, a computer generation’s length shortens over time—now about 12 months. One generation after computers match human intelligence (the 31st day), the best computers will be twice as powerful. One more year and their advantage will be fourfold. Once computers program computers, the process accelerates.

Siding with those who believe intelligence, or a reasonable facsimile of it, is in the future of non-biological systems, I believe we should begin to consider the implications. Grandma told me to learn to work with my brain because, unlike physical labor jobs, “you can never be replaced by a machine.” I chose teaching which may be a good case study for looking beyond the 30th day.

While teachers may not become obsolete any time soon, the role would certainly change. What tasks are left when the classroom has dozens of video eyes, microphone ears, and speaker voices serving a superior computer brain linked to every bit of recorded knowledge? Will schools drop academic instruction and become exclusively places of socialization? What will the curriculum be when the nature of being an educated citizen is redefined? What happens to the utilitarian, career-prep function of schooling when traditional careers are left behind?

Potential changes range from trivial to foundational. Why have Driver’s Education when cars drive themselves, react faster, see in the dark, and never get drunk or tired? More profoundly, how do you teach math when Johnny and Suzie can converse with their calculators? What math is now important? Now how do you deal with the child's question "When are we ever gonna use this?"

If computer intelligence surges past human intelligence, will job security be for those who work with their hands? The car computers may tell the garage computers the car's brakes need replaced, but a human must do the labor. That is, unless we are also in the 29th day of the age of robots.

Links to this post:

Create a Link

<< Home