The Computer Brain
By David Talbot
The neocortex constitutes the bulk of the human brain and is the presumed seat of learning, language, memory, and whatever it means to be human. It contains many billions of neurons, and each neuron can interact with nearby neurons in thousands of different ways. The operations of even a single neuron are difficult to measure, and biologists don’t agree on how many distinct subclasses of neurons are present in the neocortex, how the six layers of the neocortex interact with one another, and whether the system behaves differently from one part of the neocortex to the next.
“It’s a humongous mess,” says Michael Beierlein, a neuroscientist at Harvard Medical School. And when neuroscientists study the electrochemical processes that take place in that mess, “ultimately we just don’t know what the crucial features are, and which ones we can safely ignore: what is biological noise, what is important, what is an experimental artifact.”
Neuroscientists around the world are trying to decipher the neocortex, because understanding it better could provide insights into everything from psychiatric disorders and brain disease to learning and memory. To that end, many groups are trying to create computer models of how neurons function. A research project launched this year by IBM is the most ambitious such effort ever attempted: the company and Swiss research partners hope to create a functioning 3-D model of a two-millimeter chunk of neocortex containing 60,000 neurons–a unit known as a neocortical column.
The neuron modeling project “is going to be larger than anything done before, by an order of magnitude,” says Charles Peck, the computer scientist at IBM’s T. J. Watson Research Center in Yorktown Heights, NY, who heads the project, dubbed “Blue Brain.”
The researchers will take raw data collected from rat neurons at the Swiss Federal Institute of Technology in Lausanne and feed it into an IBM supercomputer that is among the world’s fastest. Henry Markram, the Swiss neuroscientist heading the biological end of the project, says a graphical representation of just the 10,000 neurons in a rat neocortical column will require up to two terabytes of storage–roughly the amount of data that can be held in 400 standard recordable DVDs. IBM computer scientists experienced in simulating biological systems will help build a 3-D model that mimics the interactions of these neurons and compare its performance against Markram”s laboratory data.
The job will be vast. “Think of a neuron as a tree, with roots and branches,” says Markram. “Imagine if you take 60,000 of these trees and squeeze them in the space of a pinhead. That is the kind of architecture you are looking at, with the roots of trees touching branches of other trees.” And that’s just for one neocortical column; the human neocortex is estimated to contain tens of millions of them. But if all goes well, “we will be able to see where the information goes, how it is represented, and how it is stored on a tree,” Markram says. “Then we can understand what can go wrong.” Markram believes the project could yield possible targets for drugs to treat brain diseases in 10 years.
That is certainly ambitious. “The simulation may lead to a better understanding of some of the circuitry,” says Tai Sing Lee, a computer scientist and neurophysicist at the Center for the Neural Basis of Cognition, a joint project of the University of Pittsburgh and Carnegie Mellon University. However, he adds, “Simulating the human brain and curing disease are extremely far away.” Viewed against the magnitude of the task, says Lee, IBM’s Blue Brain project is worthwhile but “a small step in biology.”