IBM has developed a revolutionary chip that is the size of a postage stamp but is equipped enough computing power to outdo a supercomputer.
The new processor, called True North, is designed to "think." That is, it simulates the activity of millions of neurons in the brain and has the ability to perform complicated tasks with just a small amount of energy. The chip does this by relying on a web of transistors that are patterned after the arrangement of nerves in the brain.
The processor, which was revealed in an article published on the journal Science, is said to be designed for multi-object detection and classification. This means that it is designed to recognize patterns the way the human brain does.
The chip, which can be powered, with a hearing aid battery, contains 5.4 billion transistors, 4096 neurosynaptic cores, one million programmable spiking neurons and 256 configurable synapses. In comparison, IBM's previous brain-like computer chip, which it unveiled in 2011, has just one core and 256 neurons.
"We have not built a brain. What we have done is learn more from the brain's anatomy and physiology," Dharmendra Modha, the manager and lead researcher IBM's cognitive computing group, said.
Modha likened traditional chips to a machine patterned after left side of the brain, which would be fast and efficient at performing mathematical computations. "What we're building is the counterpart, right-brain machine," he said.
While the new processor is being sold a massive technological breakthrough, some experts remain unimpressed. "The chip appears to be very limited in many ways, and the performance is not what it seems," Yann LeCun, an expert in neural networks who also works as Facebook's Director of Artificial Intelligence Research, said.
LeCun described the tests on the processor's ability to recognized moving cars and pedestrians as inadequate. "This particular task... won't impress anyone in computer vision or machine learning," he said.
The chip, which took more than a decade to make, is the product of a research effort called Systems of Neuromorphic Adaptive Plastic Scalable Electronics (SyNAPSE). The program obtained its $53.5 million in funding from the US Defense Department's Defense Advanced Research Projects Agency.
