Monday, March 19, 2007


*tear*

A quiet death for bold project to map the mind

Military scraps research at Rutgers and other schools to engineer a brain
Thursday, March 15, 2007
BY KEVIN COUGHLIN
Star-Ledger Staff

The military, it appears, is no match for the mind.

The Star-Ledger has learned the Pentagon quietly has killed a project to "reverse-engineer" the human brain, a goal one participant compared to inventing the atomic bomb or landing men on the moon.

Scientists from Rutgers, Harvard, the Massachusetts Institute of Technology and a dozen other schools and institutions set out in 2005 to map the complex workings of the brain, in hopes of creating software that someday might help leaner combat forces fight smarter.

Now, as researchers scramble to salvage their pieces of the puzzle, some are lamenting the loss of a rare chance to work closely with diverse experts -- a cross-pollination of neuroscientists, psychologists and computer scientists -- they say promised bold approaches to demystify the brain.

"It was all very exciting. ... It's hard to know at the outset if it was too ambitious," said Mark Gluck, a neuroscientist at Rutgers-Newark who, like others on the project, was left scratching his head over its demise.

"All we know is it's dead," Gluck said.

The program was called BICA, short for Biologically Inspired Cognitive Architectures. It was sponsored by the Defense Advanced Research Projects Agency, or DARPA, the agency that spawned the Internet.

About $9.5 million was earmarked to chart a game plan, the brain project's first phase. But the next stage -- a five-year, $50 million to $100 million push to design and test brainlike software -- never got launched.

Rumors the program had been scratched began circulating late last year among participants, who say they still have heard nothing official. DARPA's Web site still says the agency is soliciting proposals for the next phase.

But in an e-mail to The Star-Ledger, program manager David Gunning confirmed: "DARPA has decided to not pursue BICA phase 2." Requests to interview Gunning were denied.

Also scrapped was a related project, Architectures for Cognitive Information Processing, which sought to develop special computer hardware.

The brain effort linked experts from neuroscience, cognitive psychology, robotics and artificial intelligence. They wanted to replicate how different parts of the brain interact -- sometimes pulling together, sometimes not -- to solve problems. Their toolkit would include everything from imaging technologies that pinpoint brain activity to emerging theories about emotions.

"In some ways, it was like a Manhattan project or the Apollo project. Building a brain is a big task," said Randall O'Reilly, an associate professor at the University of Colorado-Boulder, who termed the program's death "pretty devastating."

O'Reilly will keep studying vision and navigation with scientists at Carnegie Mellon University, on the U.S. Navy's tab. But only DARPA, he said, is capable of supporting truly far-out, "blue sky" research among collaborators from disparate fields.

Pentagon watchers and academics have questioned whether DARPA has shifted focus, from basic research to faster programs to serve troops in Iraq and Afghanistan, and homeland security.

Anthony Tether, who took charge of DARPA in 2001 and is its longest-serving director, has maintained that his agency strives to serve both needs.

"I haven't seen any convincing data that they've tilted one way or the other," said Kei Koizumi, an analyst at the American Association for the Advancement of Science.

Tether's critics describe a micromanager who hinders freewheeling research, a hallmark of DARPA since its debut in the Cold War. Defenders hail the electrical engineer for keeping the $3 billion agency honest, for delving into biology, and for galvanizing U.S. students with robotic car contests.

Addressing Congress last year, Tether strongly endorsed so-called cognitive research, for "thinking" computers that will streamline decisions for military commanders. Computers that interpret data more intelligently -- and, maybe, reason like a human -- are considered crucial for smaller fighting forces in the future. The Pentagon hopes that automating administrative and support tasks will free up more personnel for combat. The brain project was part of this drive.

An agency spokeswoman said DARPA is trimming $300,000 from cognitive programs in fiscal 2008, but will ask Congress to boost funding down the road. One ongoing project is PAL, short for Personalized Assistant that Learns. Computers are adapting to users' quirks, to customize information in busy command centers.

DARPA officials were terse when asked why they quit deciphering the brain.

Citing unspecified "changing priorities," agency spokeswoman Jan Walker noted it's not unusual for projects to wash out.

"We are constantly starting and stopping programs," Walker said. "We review our programs every 18 months, and make decisions based on the resources available, and technical progress."

O'Reilly conceded progress would have been hard to gauge. How do you design tests to show software is learning and adapting to new situations? One idea was to train software on a video game, he said, then make it play a new game and grade the results.

"We really hoped this would be an important project and would lead to some real breakthroughs," said University of Michigan computer scientist John Laird, who had wanted to explore how emotions shape decision-making. "What was very ambitious was looking at the brain as a whole -- how the pieces fit together, and what computations could be done for the whole brain, not just a specific area."

At Rutgers, Gluck seeks other federal funds for his research on memory circuits of the brain.

But simulating the entire organ? "That's not going to happen without DARPA," Gluck said.

Kevin Coughlin may be reached at kcoughlin@starledger.com.

© 2007 The Star Ledger
© 2007 NJ.com All Rights Reserved.

0 Comments:

Post a Comment

Subscribe to Post Comments [Atom]

<< Home