Few terms carry as much emotional and technical baggage as AI (artificial intelligence). And while science-fiction authors probe AI's metaphysical boundaries, researchers are producing practical results. We may not have a robot for every task, but we do have cell phones that respond to our voice, data-mining tools that optimize vast industries, and thousands of other measurable ways AI-influenced computing enhances how the enterprise gets work done.
That said, AI itself remains elusive, and the measure of AI's position on the enterprise crackpot scale depends wholly on where you set the goals. Restricted to applying templates and well-defined theorems to sets of data with precise definitions, computers are, after all, becoming very adept at using statistics to make educated guesses about the world. And though speech-recognition software, for example, may not hear the actual message, whatever that means, it does know that a certain pattern of sounds and frequencies almost always corresponds to a particular word.
Greg Hager, a professor of computer science researching machine vision at Johns Hopkins University, says, "For a long time, people thought that the way you would solve those problems was to understand how people would solve those things and then write a program that would do what people do."
That approach has yet to produce much success, but as Hager points out, less sophisticated, more statistical algorithms that take educated guesses are becoming increasingly accurate. Some of the best algorithms for recognizing objects in images, for instance, look for salient features, waiting until enough key points are recognized. Such an algorithm could recognize a Ford sedan from multiple angles but wouldn't be smart enough to use that experience to recognize a Chevrolet.
"It's a paper-thin notion of intelligence," Hager says, but one that's still useful in many basic cases. Expect the enterprise to benefit from similar AI-inspired computing paradigms and technologies in the very near future.
-- Peter Wayner
What sort of AI-inspired technologies do you see impacting the enterprise?
Crackpot tech 2008: Crackpot technologies that could shake up IT
Eight more technologies that straddle the divide between harebrained and brilliant -- each with a promise to transform the future of the enterprise
Crackpot tech 2007: 12 crackpot tech ideas that could transform the enterprise
These technologies straddle the divide between harebrained and brilliant as they promise to shake the pillars of tomorrow's enterprise