The world's most popular search engine, Google, uses artificial intelligence to respond to millions of queries a day. Banks now depend on artificial intelligence to alert customers to odd patterns of credit card use. And many video game developers rely on AI to develop life-like characters.
After its own boom-and-bust cycle in the 1980s, the esoteric field of artificial intelligence gradually has developed some real-life uses of software that teach machines to think.
And now the war on terrorism is boosting AI research with an infusion of cash. The Defense Department hopes an elite group of AI scientists will develop more tools to help intelligence analysts find terrorists before they strike.
At an artificial intelligence conference in San Jose, Calif., last week, several groups of university researchers presented papers on work they have done in the area of counter-terrorism.
Their projects show that it is possible to use AI techniques in tracking down terrorists. But the researchers said their work was in the early stages, with some projects waiting for new rounds of funding.
"People have too much information," said Peter Jarvis, a principal software engineer at NASA Ames Research Center, describing the job of a typical intelligence analyst at the CIA or the FBI. "We developed technology to have the machine sort through this stuff."
In one research project, computers had access to multiple databases, some that had junk information, and some that contained clues for an agent to investigate. The system combed a huge pile of alerts created for the study and produced a cluster of data it believed was linked.
The system then recommended that the agent look into a series of events: a missing persons report filed on an industrial worker, a blueprint of the switches to a dam found on an al-Qaida computer, and hacking tools found on other al-Qaida computers.
"You could put that together to say that the terrorists were possibly planning to attack the dam," Jarvis said, adding that government agents could spend more time investigating actual incidents, rather than spending all their time studying data.
Jarvis added that the project was probably about five years from commercialization. The Defense Advanced Projects Research Agency provided the initial funding for the project and the intelligence community is showing an interest in the work.
"Research funding has been going up in the past two to three years," said Jaime Carbonell, a director of the Language Technologies Institute at Carnegie Mellon University.
The spike in AI funding is due in part by the need for homeland security applications, he said.
At the San Jose AI conference, a "roboceptionist" named Grace developed by Carnegie Mellon researchers showed how far the field has come and how far it has to go.
Grace was supposed to help conference-goers with directions to restaurants and other destinations. After receiving several queries with the word "restaurant", she only would respond to the phrase, "Where can I go eat?" and gave the address of the Arcadia Restaurant in the Marriott Hotel next to the convention center.
Another time, she sensed some curious onlookers. "Instead of just staring at me you could type something on the keyboard," she intoned in her computer-drone voice.
Bill Smart, an assistant professor at Washington University in St. Louis, said machine learning is the toughest task for AI researchers.
"It's still the early days," he said.
| • | Defense Department hopes to use 'AI' in terrorism war
TheState.com, Thursday, Aug. 5, 2004 Byline: Therese Poletti, San Jose Mercury News, CA |
|---|
| Story also ran in 2 others: San Jose Mercury News and CA |
|---|
|
Related Information Related Groups:
|