Updated

With artificial intelligence marking its 50th anniversary, it's tempting to recall how the technology has proven to be overhyped.

But at the American Association for Artificial Intelligence annual meeting this week, AI evangelists were touting ways that a confluence of factors could put computer learning at the heart of a smoother-running world.

One reason for uneven AI progress over the years was the vast amount of resources devoted to building "expert" systems to handle specific tasks, such as routing flights.

Not only were such systems "brittle" outside of their field of expertise, but traditionally they had to be manually taught reams of information.

However, with the help of mind-boggling increases in computer processing power, AI researchers have developed programs that can automatically glean insights from documents and detect patterns in complex systems.

Now a computer that often sees A happen before B doesn't necessarily conclude that A causes B.

Perhaps the biggest reason for today's optimism is the Web's emergence as an ever-richer storehouse of knowledge for AI programs to mine and reuse.

During a break in the conference, Microsoft Corp. (MSFT) researcher Eric Horvitz, the AAAI president-elect, illustrated by tapping a handheld computer showing a map of Seattle highways.

Combining real-time road sensor data with historical information about weather patterns, traffic flows at certain times of day and even pro sports schedules, the color-coded map was indicating how long drivers could expect to wait before gridlock hit each artery.

The program can signal when drivers would be surprised, pleasantly or unpleasantly, by conditions on a certain road.

That Microsoft program, known as JamBayes, is one of many aimed at turning computers into discreet assistants. A researcher at SRI International demonstrated a program that can schedule meetings, delegate tasks and book trips for a harried office worker, asking for guidance from the human when conflicts arise.

Horvitz said he envisions a future of people and computers dancing an information fox trot together: You do your part, I'll do mine.

"There is an elegant path to doing this well," he said. "We're trying to unlock the expected utility — or happiness — of people."