Quantcast

New IBM-Developed Processor Functions Like The Human Brain

August 9, 2014
Image Caption: The project, which was funded by DARPA, could allow a chip to perform supercomputer-level calculations without needing to connect to the Internet to do so. Credit: Thinkstock.com

redOrbit Staff & Wire Reports – Your Universe Online

IBM researchers have announced the development of a new computer chip that is inspired by the brain, mimicking the way that the mind can recognize patterns utilizing a web of interconnected transistors to simulate neural networks.

The processor is named TrueNorth, and according to John Markoff of the New York Times, it contains more than 5.4 billion transistors, yet requires no more power to function than a hearing aid (just 70 milliwatts of power versus the minimum of 35 watts required by current Intel processors, with have 1.4 billion transistors).

TrueNorth contains electronic “neurons” capable of signaling others when a specific type of data reaches a predetermined threshold, allowing them to work in unison to organize data into patterns, Markoff said. Using this infrastructure, the chip could ultimately be capable of calculations beyond the modern supercomputer, recognize when a person is performing a specific action, or controlling the activities of a robot.

Despite being no larger than a postage stamp, this neurosynaptic processor could also be used in self-driving vehicles and artificial intelligence systems installed on mobile devices, the AFP news agency explained. It is part of the company’s new approach to computer architecture design known as “cognitive computing.”

“We have taken inspiration from the cerebral cortex to design this chip,” IBM chief scientist for brain-inspired computing Dharmendra Modha told the news agency. He and colleagues from Cornell University and Cornell Tech detail their findings in the latest edition of the journal Science.

Modha explained that the computers we use today trace their lineage back to the “sequential number-crunching calculators” of the 1940s, with focus solely on mathematical or “left brain” processes. TrueNorth, on the other hand, attempts to mimic “right brain” functions of sensory processing by responding to visual, olfactory and other stimuli in order to learn how to respond to different situations, AFP added.

The project, which was funded by the US Defense Advanced Research Projects Agency (DARPA), could allow a chip to perform supercomputer-level calculations without needing to connect to the Internet to do so. This would allow autonomous cars to detect and solve potential accidents and other problems without needing to have a connection to Wi-Fi, and smartphones to interpret sights and smells in real time and without the need for a data connection.

“Though it is providing few details on timing, IBM says it is already talking to potential partners about ways to bring the chip to market,” said Wall Street Journal reporter Don Clark. “The company has connected multiple chips together to test potential system designs, and sees applications of the technology ranging from room-size supercomputers to floating jellyfish-shaped devices that could sense tsunamis or other aquatic conditions.”

Modha added that the company has “huge commercial ambitions” for the technology, but emphasized that the neurosynaptic chip is “not going to replace conventional computers. It is a complementary relationship.” While Intel, Qualcomm and others are conducting similar research, Cornell Tech electrical and computing engineer and project collaborator Rajit Manohar said that TrueNorth is “much closer to being usable” than their competitors’ processors.

Image 2 (below): A circuit board shows 16 of the new brain-inspired chips in a 4 X 4 array along with interface hardware. The board is being used to rapidly analyze high-resolutions images. Courtesy: IBM

—–

Shop Amazon – Hot New Releases – Updated Every Hour


Source: redOrbit Staff & Wire Reports - Your Universe Online

New IBM-Developed Processor Functions Like The Human Brain


comments powered by Disqus