As Computers Grow Ever Smarter

The conventional wisdom says machines are great calculators but not creators of new ideas. That belief may soon be eclipsed

By Justin Kovac

Computer-science researchers have been discussing the prospect of machines that are smarter than humans for years. They don't mean just that computers can juggle numbers far faster than any human mind. They envision computers that surpass humans in ingenuity and creativity.

How do you program creativity and ingenuity? You can't -- not with traditional step-by-step programming, at least. However, other approaches are possible.


  Consider the hallmarks of what we recognize as ingenuity. It's basically the ability to put information together in new and unexpected ways. Given all the scientific and technological information that's now available online, and the speed at which computers could assemble and test multitudes of combinations, it's likely that computers will soon surpass human ingenuity.

Even if their batting average never comes close to that of clever people, the productivity of ingenious computers will surely outstrip that of most humans by virtue of sheer speed and brute-force processing.

Creativity, on the other hand, would seem to be utterly different. It is as immeasurable as it is invaluable. Creativity is not the product of a certain number of man-hours, so adding processing speed does not help. Creativity involves developing new ideas -- original and without logical precedent. You might think that a computer would have trouble finding solutions to problems that can't be described with logic. But biology provides a path.


  A recent trend in computer science is called "genetic programming." It emulates evolution. Small pieces of data (genes) are randomly assembled into several different models (organisms). The models are tested on a problem to determine how close they come to finding a solution (viability). The models that perform well (the fittest) then mix with each other (mate) and produce hybrid models -- and a new round begins. The new generation consists of the best parents from the previous, their progeny, and a few randomly generated models to mimic mutation.

After thousands or millions of generations, this artificial evolution should home in on the optimum solution. While this process is limited somewhat by the number of parts at the start and the rules used to select the most promising candidates, a computer can evaluate far more parts and combinations than any human. In actual applications, genetic programming often evolves original solutions that surprise even the most astute human experts.

As computers advance in speed and complexity, some scientists believe the technology won't stop with exceeding human ingenuity and creativity. The Matrix and I, Robot are works of science fiction, but the questions they raise may soon take on a real-world context: If we can create machines that are more creative and more ingenious than humans, can we create a machine endowed with free will?


  Today's computers are slaves to the data we humans provide to them. So, when a computer is powerful enough to develop a new economic model, one better than people can craft, humans will eagerly supply the programming. But when a computer can independently recognize the need for a better economic model and autonomously chose to develop it, the implications would be profound.

Sharing this planet with another sentient species is something that most people have never seriously contemplated. Should we create a race of potentially superior beings? Unless something slows the progress in computer power, we will soon have to decide.

Kovac was a finalist in the 2005 Intel Science Talent Search

Before it's here, it's on the Bloomberg Terminal.