Of Bugs & ’Bots

Uncommon Ground - Of Bugs & ’Bots - Spring 2006

Modern technology produces some impressive machines that can ?nd their own way down a narrow corridor or play a game of chess, but even the most sophisticated arti?cial intelligence can’t come close to matching the cognitive abilities of a bee. But according to Fred Dyer ’77, a zoologist at Michigan State University, the intelligence of an insect is exactly what engineers should be striving for in future robot generations.

More speci?cally, Dyer—who spoke to Denison students last year about the potential relationship between robots and insects—and his fellow researchers hope that studying biology will reveal secrets to building a robot that can learn from its environment without receiving outside instruction and learn to not repeat errors, just like insects do. It won’t be easy. Dyer said there’s a misconception that any brain smaller than ours has generally diminished capacities. In truth, many creatures, even insects, have speci?c cognitive abilities that are superior to our own. “Insects can’t think about poetry or chess, but they don’t need to,” he said. They do need to know how to learn and navigate new and changing environments, and they excel at these tasks.

Presumably insects’ brains are among the simplest in the animal kingdom and the best place for arti?cial intelligence engineers to start. “But even with insects, there are a lot of deep mysteries,” said Dyer. An insect’s ability to investigate an anomaly and report its ?ndings to others of its kind continues to confound technology researchers, said Dyer.  

In his own research, Dyer documented how a bee, when released into a large circular container with a food source at the center, makes a “beeline” directly to the food and goes out the way it came. But, if an object is placed anywhere near the bee’s path, it will interrupt its ?ight and buzz around the object to determine if it will be a useful landmark for return trips. Afterward, the bee will ignore the object and return to its original ?ight path until it encounters something new. No machine is capable of such ?exible responses to landmarks today nor is one likely to be soon.

“The idea of a fully functional robot with one millionth the mental capacity of an ant is still really far off,” said Dyer.

Published March 2006