06/23/97 EDINBURGH: WHERE ERSATZ CRICKETS CHIRP

Return to main story

EDINBURGH: WHERE ERSATZ CRICKETS CHIRP
Its ''informatics'' unit fuses AI, computers, and cognitive science
A psychedelic poster for an artificial-life conference is taped to the wall. Colorful balls used in an experiment dangle from the ceiling. Some of the Generation Xers working at the University of Edinburgh's large mobile- robots lab pad around in bare feet and shorts. And in the center of it all, Henrik H. Lund, a 27-year-old Danish researcher sporting a long blond ponytail, is playing love songs to his ''crickets.''
He's not romancing real bugs. He's wooing tiny robots with microphone ears. They were built to demonstrate how crickets in nature find their singing mates. If he claps, they won't budge. But sure enough, when Lund turns on a speaker playing the recorded songs of real crickets, one of the robots sashays toward it.
NOBLE PEDIGREE. Working with Barbara Webb, an animal behaviorist, and John C.T. Hallam, director of the mobile-robots lab, Lund has replicated the parts of the cricket's hearing and nervous system that help it navigate. In effect, the team built a software and hardware model of a real cricket. ''This is as much about biology as it is about artificial intelligence,'' Lund explains.
As Lund's crickets show, boundaries between the sciences are blurring at Edinburgh. Founded in 1583, this university is a hotbed of ''informatics''--a discipline that unites computer science, artificial intelligence, and cognitive science. The goal is ''developing a new science of information,'' says the division's head, Michael P. Fourman.
The work builds on a notable pedigree. The predecessor to the university's current AI department was set up in 1963 by Donald Michie, who helped crack the Nazi's Enigma code during World War II. In the 1970s, the department invented the Prolog programming language. Until Cambridge University recently outdid it, Edinburgh boasted Europe's most powerful supercomputer. Today, Edinburgh is one of the few research groups outside the U.S. to get support from the Defense Advanced Research Projects Agency, America's chief funder of AI research.
Closer to home, the university just developed ''intelligent agents'' that help companies such as Unilever PLC and IBM UK manage work flows. The resulting software, explains B. Austin Tate, technical director of the AI department's technology-transfer office, does everything from guiding companies' bids on contracts to automating big chunks of business operations.
Already, one of Edinburgh's inventions helped run a small part of America's most famous business school. Harvard tested an Edinburgh program that used so-called evolutionary algorithms to manage exam schedules. Edinburgh's research is also helping a Scottish company, Marshall Agriculture, schedule the collection and slaughtering of more than 1.3 million chickens a week.
Edinburgh's computer finesse spills over to the social sciences. It is home to some of the leading thinkers on the relationship between men and machines. One, sociologist Donald MacKenzie, collaborates with Edinburgh's informatics team. MacKenzie has written an award-winning book on such questions as how the design of ballistic missile guidance systems in the U.S. and Russia reflects the structure of their respective military organizations.
It's easier to analyze the man-machine interface than to improve it. But one informatics team is trying. Bringing together AI researchers and cognitive scientists, the team has created what it calls an intelligent labeling explorer, or Ilex. Working with the National Museums of Scotland, this team created a prototype that provides a virtual tour of a 20th century jewelry exhibit. Unlike museum labels, Ilex tailors the tour to the interests of the person being guided. For example, if you linger on a Coco Chanel bracelet, the system might suggest that you look at another French designer working in the same style.
To make machines more human, a bit of humor helps, researchers say. So one group is schooling computers in puns. The machines don't giggle, but they seem--in their way--to grasp why crossing a monkey and a peach might yield an ''ape-ricot.'' The researchers haven't tried that one out on the crickets yet. All in good time.
By Julia Flynn in Edinburgh

Return to main story