A couple weeks ago, I was invited to see the new Rolex Learning Center at the Ecole Polytechnique Fédérale de Lausanne. Architecture critics are fawning over the $65 million building. A writer in the Guardian yesterday compared it to "some filmic version of the afterlife." But to me what's even more remarkable is what's going on nearby at the school: they're building a brain.
The Ecole Polytechnique Federale de Lausanne is a grim campus of sparsely-windowed gray buildings thirty minutes north of Geneva, Switzerland. When I was last there a couple years ago, I met Marcus Bartschi, a wry, 43-year-old computer engineer, who sat at a desk in a dim office with a rusty saw hanging inexplicably on a bare concrete wall. He had a scruffy black beard and dark circles under his eyes, brought on by his two-week-old son.
“My boy has only physical needs now,” Bartschi told me, resignedly, “there’s not much for me to do.” So he tended to his other baby: a four ton supercomputer named Blue Gene. Developed by IBM, Blue Gene is one of the world’s most powerful high-performance machines, and it was being used here for a suitably super, some say impossible, mission: to simulate a brain. Bartschi had the vulnerably human task of keeping it alive. “Hellooooo,” he cooed, as the laptop monitoring Blue Gene fires up.
The project, dubbed Blue Brain, is the dream of Dr. Henry Markram, the neuroscientist in charge of the EPFL’s Brain and Mind Institute.Several years ago, Markram approached IBM with an ambitious plan: to accelerate the normal path of research by building a three-dimensional model of a mammalian brain, as he said, “in silico.” Skeptics such as MIT’s Marvin Minsky are already questioning what they call won’t rule it out completely. “We do not yet know if consciousness will emerge in these artificial brains,” he has said, “and we will consider the ethics of this if this happens.”
Bartschi, a diehard Hitchhiker’s Guide fan, was more skeptical about his Blue Gene baby, “this is no HAL,’ he saidBlue Gene, however, was robust (with a processing speed of 22.8 tera-flops per second, the eight fastest supercomputer in the world) and affordable ($2 million per rack) enough to give Markram his crack. The goal was to first simulate a single neocortical column, then take another ten or so to replicate an entire brain. While Markram’s team tends to the data-crunching software, Bartschi had the less glamorous job of caring for Blue Brain’s pillar: Blue Gene. “It’s a lot of pressure,” he said, with a sigh, “I want it to work 100% of the time.”
After the fanfare of the announcement, Bartschi oversaw the installation of the four refrigerator-size racks, which resembled the black monoliths from 2001: A Space Odyssey.After two weeks of assemblage, he triumphantly booted up Blue Gene – only to see it repeatedly, and mysteriously, shut down. For two weeks, Bartschi searched for an explanation, only to discover that a solid steel beam underneath the raised floor was obstructing 50% of the airflow required to keep it cool. It took another two weeks for Bartschi’s team of six to schlep Blue Gene down the hall. The next month, a crucial fan broke, and there was no replacement on site. The part was ordered from Rochester, New York, but, due to a mix-up in shipping, was sent low priority – holding up Blue Gene for more than a week. Despite such tantrums, Bartschi had developed a soft spot for the machine. “The whole setup is complex," he said, “In this sense, it is a living thing.”