Friday, December 19, 2008

Computer Smarts

In 1965 Gordon Moore observed that for nearly two decades computer circuitry had gotten miniaturized at an incredible rate: every two years the number of transistors that could be put into an integrated circuit had roughly doubled. He predicted that this trend would continue for some time in the future.

And it has, for nearly half a century now.

This is remarkable--it means computer capacity has doubled 22 times. If you had planted a pea plant one inch high in 1965 and it had sustained this growth rate, it would now tower well above the stratosphere. If you had wisely invested one dollar at this rate in 1965, the value of your investment would now be over $5 million.

In addition, computer innards have gotten more and more cleverly linked together (thanks to advances in computer design and programming), and relationships among computers have gotten faster and more extensive (thanks to networks within a room or a building, or even more widely like the Internet and the World Wide Web).

Computers have gotten quite a bit cleverer, too. They used to think algorithmically, that is, step by step like a cooking recipe. But they also learned to think heuristically, that is, using generalizations and approximations and rules of thumb. Moreover, with the development of a field called "artificial intelligence," they became increasingly able to learn from experience, that is, to try something and if that didn't work, to try something else (and to remember not to try the first method again in tackling future problems).

Chess is probably the hardest, most complicated game around. In the 1960s, '70s, and '80s, chess-playing computers emerged from their kindergarten to become more and more competitive against human players. By the 1990s they were tying and sometimes beating chess grandmasters; nowadays they regularly beat the greatest grandmasters in the world.

But there are still some mental activities that the largest, best, fastest computers cannot do as well as humans. These are particularly visual processing, spatial reasoning, and some forms of problem solving. In fact, there are some websites designed to get people to help out computers (and the scientists and business folks they work for).

For example, http://www.galaxy.org gets some 160,000 home computer users to help classify images of millions of galaxies that have been photographed by giant telescopes, and to find patterns and oddities. This has been remarkably successful at spotting curiosities that computers might not have noticed.

Another website, http://fold.it, which now has over 60,000 home participants, presents difficult puzzles in 3-D folding of protein molecules. Several remarkable discoveries on the complex structure of proteins have emerged from participants work (or "play") at this website.

Google has set up a website game, http://images.google.com/imagelabeler, to get human observers to help find the best, most recognizable names for a wide variety of pictures--of birds or wildlife, of industrial or city scenes, or of abstract or obscure objects, for example.

And finally, http://www.gwap.com, provides "games with a purpose" to help train computer vision and artificial intelligence systems. It has some 120,000 user/participants.

Computers have grown up fast over the past few decades. They have revolutionized our lives and continue to rush forward, creating the future at an accelerating rate. Every generation should exceed the abilities and accomplishments of its parents; our silicon offspring have overtaken us and left us standing by the roadside in many ways. But there are still ways in which they need our help.