Today, a tersely worded note arrived in my inbox from an IBM spokesperson, who said, "IBM Research completed the first phase of the BlueBrain project and we are not involved with the second phase."
This was something of a surprise, given the hype surrounding the future of the Blue Brain collaboration, and the extensive roadmap that had Henry Markram, the project's director, building ever bigger and more complex models on ever more capable Blue Gene upgrades. The final goal was the full-scale, near-biological simulation of a human brain.
The 22.8 Tflop IBM Blue Gene/L currently humming away in the basement of Switzerlandâ''s Ecole Polytechnique Federale de Lausanne houses 10 000 neurons of a virtual ratâ''s neocortical column, a microcircuit that is in reality about the size of a fine pen tip. Markram told me that the next phase of Blue Brain would involve ganging these neocortical columns together in either a visual cortex or a somatosensory cortex. For that, he said, he would need more computer than the current, 8092-processor machine was capable of.
He also mentioned that IBM would be sending him a new machine in January 2009.
As recently as mid-December, the project was still on as far as IBM's PR people were concerned. "The Blue Brain project is a comprehensive attempt to understand brain function and dysfunction through detailed simulations," an IBM public relations minister told me proudly in an email likely lifted from a press release. "Analogous in scope to the Genome Project, the Blue Brain will provide a huge leap in our understanding of brain function and dysfunction and help us explore solutions to intractable problems in mental health and neurological disease."
It would be tempting to blame this on the recession, but IBM just hopped on board the DARPA train to spearhead the SyNAPSE project, which is computational neuroscience-- a different ball of wax from Blue Brain. Here's a detailed explanation of the differences.
If anyone has any insights into what happened there at Big Blue, please email me.