The US Air Force Research Lab and IBM research is planning to construct and deliver a powerful supercomputer running on the brain inspired IBM TrueNorth Neurosynaptic System. This scalable platform that IBM is building for the air force is going to feature an end to end software ecosystem which is designed in order to allow for deep neural network learning as well as information discovery.
The advanced pattern recognition system and sensory processing means is going to be equal to 64 million neurons while the processing portion consumes energy equivalent to a dim light bulb.
IBM TrueNorth Neurosynaptic System to be more efficient
According to IBM researchers, the brain inspired neural network design for TrueNorth is going to be more efficient as concerns pattern recognition and integrated sensory processing compared to the systems which are enabled by conventional chips.
It seems the Air Force Research Lab is investigating applications for the system within embedded, autonomous settings where at the present, the weight and power and crucial factors that limits production.
Both the USAF and IBM announced today the machine is going to run on an array of TrueNorth Neurosynaptic chips. These TrueNorth chips are apparently wired and made in a fashion which resembles the synapses within the brain of an individual.
Each of the cores is going to be part of a distributed network and operate parallel to each other on an event driven setting. As such, these chips do not come available with a clock the way conventional central processing units do in order to function.
Similarly because of the distributed nature of the system, even if one of the cores shuts down, the rest within the array continue. As with other neural networks, the system is going to be operated within pattern recognition and sensory processing states.
The Air Force is trying to combine the ability of TrueNorth in the conversion of multiple data feeds, whether it is video, text or audio into the machine readable symbols with the conventional supercomputer ability when it comes to processing information.
This neurosynaptic is going to be utilized in order to allow for new computing capabilities which are crucial to the mission of the Air Force Research Lab and its mission to explore and demonstrate high impact and game changing tech which is going to allow the Air Force to maintain technical advantages.
A single processor within the system is consistent of 5.4 billion transistors that are set into 4096 neural cores which create an array of 1 million digital neurons communicating with each other via 256 million synapses.
The IBM TrueNorth Neurosynaptic system was first created under the auspices of DARPA’s SyNAPSE program while partnering with Cornell University. Though it is not the first time the IBM neural chip system has been integrated into upcoming technology.
In the previous year, Samsung’s installed chips within the Dynamic Vision Sensors thus enabling cameras to capture images at 2000 fps while utilizing 300 milli-watts.