mirror from:

DARPA project will study neural network processes

October 26, 1998
By William Jackson

A California software company has received $3.3 million from the Defense Advanced Research Projects Agency to build a cortronic brain—a phrase that recalls science fiction writer Isaac Asimov’s positronic brain.

HNC Software Inc. of San Diego will spend the seed money to develop cortronic neural networks conceived by its chief scientist Robert Hecht-Nielsen.

Hecht-Nielsen said the research could revolutionize understanding of the human brain and improve artificial intelligence techniques. “We think we have got the universal computing process,” he said.

DARPA was “sufficiently interested to help him flesh out his ideas a little,” said Ronald L. Larson, assistant director of Intelligent Software and Systems in DARPA’s Information Technology Office. “Any future support at this point is unclear.”

The Office of Naval Research is administering the project. If any valuable technology results, the government will have free use of it, while HNC Software will retain commercial rights.

Hecht-Nielsen said he is building on the work of Karl Steinbuch, a German professor who in the 1950s proposed the lernmatrix to explain how the brain’s firing neurons follow established patterns in processing new types of information.

Although Steinbuch’s ideas have never been fashionable, “I’m pretty sure they’re not wrong,” Hecht-Nielsen said. “This should be controversial, but it ought to generate a lot of interest.”

He plans to build a lernmatrix of software loaded on a PC network. His cortronic system will try to distill a core of specific rules from databases of text, visual and audio data that it can apply to new situations.

“This is going to require a lot of computation,” Hecht-Nielsen said. “It will take a long time to prove our point, but we’ve got enough experimentation already done to make some progress.”

For DARPA, whose charter to explore new technologies brought about such advances as the Internet, a cortronic system could relieve a growing bottleneck.

“We’re wrestling with the same problem as everyone else,” Larson said: “how to deal with all the information we’re getting.”