IBM's New Microchip Simulates Human Brain
The chip represents a major advancement in emulating brain functions using silicon.
IBM has developed a microchip that performs calculations in a similar fashion to that of neurons and synapses in the human brain. Named TrueNorth, the chip is designed for low power consumption, and can recognize patterns and classify objects at a more efficient rate. This represents the latest success in a $3 billion pledge for semiconductor research.
TrueNorth was built by Samsung through the same manufacturing methods used to make microprocessors for smartphones and other mobile devices. The basic design was the result of a collaborative effort between IBM and researchers at the New York campus of Cornell University, and said project has been given $53 million in funding from DARPA since 2008.
The chip uses 5.4 billion transistors (more than four times the number in a conventional PC processor) to generate the equivalent of one million neurons and 256 million synapses. These transistors are organized into 4,096 structures called "neurosynaptic cores"- each able to store, process and transmit data to each other via a communications scheme called a crossbar. This crossbar also allows for TrueNorth to be considerably more energy efficient that normal microchips; instead of drawing 50 to 100 watts per square centimeter, it only draws 20 thousandths of a watt.
IBM is already searching for potential business partners to help in bringing the chip to the consumer market. "We have huge commercial ambitions," said Dharmendra Modha, an IBM researcher and chief scientist for brain-inspired computing. Possible applications for the TrueNorth chip vary from room-sized supercomputers to aquatic devices that could sense changes in oceanic conditions. They could even be used in rolling robots with cameras that could inspect areas after a natural disaster.
Share your thoughts about this story in the comments!
Source: The Wall Street Journal [http://online.wsj.com/articles/ibm-unveils-chip-simulating-brain-functions-1407434402]
Permalink
The chip represents a major advancement in emulating brain functions using silicon.
IBM has developed a microchip that performs calculations in a similar fashion to that of neurons and synapses in the human brain. Named TrueNorth, the chip is designed for low power consumption, and can recognize patterns and classify objects at a more efficient rate. This represents the latest success in a $3 billion pledge for semiconductor research.
TrueNorth was built by Samsung through the same manufacturing methods used to make microprocessors for smartphones and other mobile devices. The basic design was the result of a collaborative effort between IBM and researchers at the New York campus of Cornell University, and said project has been given $53 million in funding from DARPA since 2008.
The chip uses 5.4 billion transistors (more than four times the number in a conventional PC processor) to generate the equivalent of one million neurons and 256 million synapses. These transistors are organized into 4,096 structures called "neurosynaptic cores"- each able to store, process and transmit data to each other via a communications scheme called a crossbar. This crossbar also allows for TrueNorth to be considerably more energy efficient that normal microchips; instead of drawing 50 to 100 watts per square centimeter, it only draws 20 thousandths of a watt.
IBM is already searching for potential business partners to help in bringing the chip to the consumer market. "We have huge commercial ambitions," said Dharmendra Modha, an IBM researcher and chief scientist for brain-inspired computing. Possible applications for the TrueNorth chip vary from room-sized supercomputers to aquatic devices that could sense changes in oceanic conditions. They could even be used in rolling robots with cameras that could inspect areas after a natural disaster.
Share your thoughts about this story in the comments!
Source: The Wall Street Journal [http://online.wsj.com/articles/ibm-unveils-chip-simulating-brain-functions-1407434402]
Permalink
