IBM Simulates A Cat's Brain -- With 147,000 CPUs
replicated the cerebral cortex of your average house cat
At the SC09 supercomputing conference in Portland, Ore., this week, Big Blue announced that it has made "significant progress" toward creating a computer that simulates a living organism's brain with abilities of sensation, perception, action, interaction and cognition. Best of all, perhaps, is that IBM said such a computer system could rival "the brain's low power and energy consumption and compact size."
Specifically, scientists from IBM Research, along with the Lawrence Berkeley National Lab, have reportedly performed the first "near real-time cortical simulation" of a cat brain that contains 1 billion spiking neurons and 10 trillion individual learning synapses. The simulation was performed on Lawrence Livermore National Lab's Dawn Blue Gene/P supercomputer, which has a whopping 147,456 CPUs and 144 terabytes of main memory.
That may seem like a lot of hardware to replicate a small animal's brain, but considering the computer has to make trillions of firing synapses across a billion neurons, IBM needs all the computing power it can get. But even with all that hardware, the simulation isn't a true artificial intelligence. The computer is merely tracing how a cat's brain activity flows and how its thoughts are formed, and the simulation is running at a far slower pace that those thoughts occur in a living animal " about 100 to 1,000 times slower.
The computer doesn't act like a cat, either (nice to know the researchers don't have to deal with shedding, clawing, laziness, and generally selfish behavior in the lab).
Thus, the computer simulation is an extremely slow brain that can't replicate an animal's behavior. So there's no need to panic about a forthcoming invasion of Cylon felines or start waving around Bill Joy's well-traveled essay about the dangers of artificial intelligence and advanced robotics " at least not yet.
But IBM's new computer is seen as a tremendous, even unprecedented, achievement in computing. And IBM isn't done, either. The research team suggests that a full-scale computerized simulator of a human brain could be attainable in the very near future. According to the project's abstract, "with further progress in supercomputing, real-time human-scale simulations are not only within reach, but indeed appear inevitable."
To that end, IBM also announced the development of a new algorithm, in collaboration with Stanford University researchers, that can measure and map the connections of all cortical and sub-cortical areas within the human brain. IBM says such mapping is crucial to understanding how the brain processes information.
"As the digital and physical worlds continue to merge and computing becomes more embedded in the fabric of our daily lives, it's imperative that we create a more intelligent computing system that can help us make sense of the vast amount of information that's increasingly available to us, much the way our brains can quickly interpret and act on complex tasks," said Josephine Cheng, IBM Fellow and lab director of IBM Research, Almaden, in a press statement.
The project was funded by the U.S. military's Defense Advanced Research Projects Agency (DARPA). IBM and its university partners received $16.1 million for phase 1 of DARPA's Systems of Neuromorphic Adaptive Plastic Scalable Electronics (SyNAPSE) initiative.