Researchers in Switzerland are collaborating with IBM to build a brain inside of a computer. The Blue Brain, a variant of IBM’s Blue Gene supercomputer, can handle 22.8 trillion operations per second. These operations will be used to model the biological behavior of individual neurons, modeling a brain from the bottom up.
Begun in 2005, the project has just completed its feasibility phase. Most of its critics fell silent when they were able to simulate a rat’s neurocortical column. This structure is one of the repeating functional units of the neocortex. The model contained 10.000 neurons with 30 million synaptic connections.
This reductionist approach differs from the functional models used by most researchers in artificial intelligence (AI) and computational neuroscience. “What they typically do is begin with a brain function they want to model and then try to see if they can get a computer to replicate that function,” says Henry Markof, director of the Blue Brain Project. He continues, “These models might help us think about the brain, but they don’t really help us understand it. If you want your model to represent reality, then you’ve got to model it on reality.”
The computational demands for this model are incredibly high. A single neuron requires 400 independent simulations. A rat’s brain has about 200 million neurons. Compare that to a human brain which has between 50 and 100 billion neurons, or 500 times that of a rat.
Markham hopes to have a rat brain modeled in a few years, a human brain in ten. It may be both sooner and later than that. Ray Kurzweil observed that scientists are often too optimistic in the long run, but too pessimistic in the short run, In other words, short term projections are often too long, and long term projections too short.
Blue Brain Project at IBM
Article at Seed Magazine
Article at Technology Review
