Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Mr. TOP SECRET]

<a href="http://www.youtube.com/watch?v=ZABeQ5vkpXM" target="_blank">http://www.youtube.com/watch?v=ZABeQ5vkpXM</a>

By Madhumita Venkataramanan
Wired.com
April 29, 2011

The brain is our body’s natural multi-system parallel processing organ. Its job, on a continuous basis, is to compute a huge onslaught of incoming data and spit out energy-intensive outputs— keen color vision, a range of auditory faculties, creation and preservation of memories.

The Army needs a low-power machine for pretty much those exact reasons. “Massive computational resources [are] necessary,” says the Army solicitation, “to support soldier-wearable computation, such as digital imaging, acoustic comprehension, and other power intensive applications.” The idea is one of a number of calls for research, made by the Department of Defense this month.

So the Army is looking for researchers to start work on a powerful computer processor, modeled to work exactly like the human brain: a “neuromorphic parallel processor.”

Using the electrochemical signals by which neurons communicate, the research will try to harness the power of the intricate neuronal network, to create a multi-parallel system to “augment processing capabilities and decrease power necessity”.

Like the brain, this processor will consist of individual neuron-like elements that will have dynamic plasticity—the ability to reconfigure their connections in response to any feedback. When this happens in our brain, it’s called learning.

The idea isn’t a stab-in-the-dark. Academic work on modeling and emulating internal brain signals has been extensive. In fact, Kwabena Boahen at Stanford University has already developed an exemplar of a hardware simulation of neuronal behavior, what bioengineers call “neuromorphic architectures.” The problem with it is that it is much too energy-demanding.

The first step in this project would be creating a practical design for a machine, by pooling together all the current knowledge on neuromorphic research. The components will be molded along the lines of biology and the goal of this stage is to have multiple “neuron” units communicating simultaneously and effectively.

In phase two, the objective is to have a hardware prototype of the machine built. The visualized specs are a “processing power of 50 million neurons in a package that weighs less than 20 grams, occupies less than ten cubic centimeters and operates on less than two pico-joules per operation.”

The final product will self-adjust its signal strength based on outside stimuli and remember its adjustments after being powered off. The Army foresees it will be used for missions where minimal power consumption is essential; for example, surveillance, reconnaissance and in-flight targeting of precision weapons.