IBM's recently announced TrueNorth chip follows the principles of brain-inspired neuromorphic computing, which aspires to intertwine computation and memory in the manner of an organic neural network.
Among the challenges to achieving this milestone is the fact that not enough is known about neuronal computing yet to design a device in which algorithms are embedded in the hardware, says Massachusetts Institute of Technology professor Nir Shavit. He says neuromorphic computing researchers are building a platform for running machine-learning processes faster and with greater efficiency.
TrueNorth is unique in that it can be tiled into an array, with chips positioned next to each other in any configuration able to communicate without additional hardware. In addition, a key advantage of the chip is its energy efficiency: it can perform 46 billion synaptic operations per second while consuming only 70 milliwatts.
IBM's Dharmendra Modha, recipient of the 2009 ACM Gordon Bell Prize, says one of his goals through projects such as TrueNorth is to embed intelligence in network-edge sensors. "Truly, our vision is a cognitive planet," he notes.
A brain-like chip also has implications for artificial intelligence (AI), although University of Maryland professor Jennifer Golbeck is uncertain of its impact at this point. Golbeck says TrueNorth could be beneficial for AI uses involving pattern recognition, such as self-driving automobiles.
View Full Article
Abstracts Copyright © 2015 Information Inc., Bethesda, Maryland, USA
No entries found