HERE COMES SANTA CLAUS - AND HE’S BRINGING ROBOTS

FAYETTEVILLE, Ark. - Robots are headed for your house. U of A robotics researcher Doug Blank predicts that this Christmas season parents will bring home robots in unprecedented numbers, and many of them won’t even realize it.

"This will be the Christmas of the robot," said Blank, assistant professor of computer science and computer engineering. "Consumers will be faced with a huge number of new robotic toys. Whether it is an adorable teddy bear or the season’s must-have toy, these are true robots."

Blank knows his robots. One of his robots took second place this year in the national robotics competition sponsored by the American Association for Artificial Intelligence. He is also the creator of XRCL, the open-source robot control language.

This year’s toys will differ dramatically from the old "press my paw and I’ll play a recorded message" toys of years past. They will possess artificial intelligence, so they can learn and respond to an outside stimulus.

These toys will generally possess what Blank calls the traditional, "top-down" artificial intelligence. In this approach, the programmer decides on a behavior and then provides the robot with a set of rules that should produce that behavior.

For example, if the programmer wants the robot to be able to move around a room independently, the robot would be given rules like "this is a wall" and "don’t get close to the wall." The robot processes these rules to produce a behavior - it moves around without running into walls. But sometimes it might go out the door and down the street.

"It is like raising a child," Blank explains. "You give the child some rules and let it go. The child may follow all of the rules and still produce very unpredictable behavior." In fact, one of the primary definitions of artificial intelligence is the science or art of making a machine that has intelligence.

Blank has pioneered a different approach to artificial intelligence, which he explains as a "bottom-up" approach. Blank presented his concept of this approach - called emergent artificial intelligence - last week at the Midwest Artificial Intelligence and Cognitive Science Society Meeting.

"This approach allows the intelligence to emerge from an interaction of many small, relatively stupid pieces of the system," said Blank. "The robot's behavior is a direct result of this interaction."

Rather than programming in a set of rules, Blank's approach provides the robot with a lot of discrete pieces of information. The robot draws on that information to make decisions and take actions within its physical capabilities.

"The system is self-organizing," he explains. "When confronted with a challenge, the robot processes the information and determines an appropriate course of action. If it has enough information, it will succeed, but it may be in a very different way than I would have thought of, even though I built and programmed the robot. This is where we will get genuinely intelligent systems."

This approach to robotics also requires a new way of thinking about what is recognized as intelligence. Blank argues that at least half of what humans call intelligence is really in the eye of the beholder. The system is only as intelligent as the complexity of the concepts needed to explain its behavior.

"For example, moving forward is a very concrete concept and requires almost no intelligence to explain," he says. "However, rationality requires far more complex concepts to explain. A robot that can make rational choices, then, is far more intelligent than a robot that can just move forward."