New-Fangled Computer Chips Could Mimic Brain Synapses
Jun 27, 2011 6:00 AM PT
A recently published University of Exeter paper brings the possibility of achieving brain-like computing one step closer.
The paper, published in Advanced Materials, states the result of a study that demonstrated for the first time the ability to simultaneously perform information and storage tasks using phase-change materials (PCMs).
The PCMs were used to reliably execute the four basic arithmetic functions -- addition, subtraction, multiplication and division -- and store information.
PCMs are substances with a high heat of fusion that can store and release large amounts of energy when they are melted and solidified at particular temperatures.
This indicates that phase-change materials can be used to manufacture artificial neurons and synapses, the researchers said.
The University of Exeter did not immediately respond to TechNewsWorld's request for additional comment.
Some Facts about the U. of Exeter's Work
The University of Exeter's experiment used a chalcogenide alloy of germanium, antimony and tellurium that's known as "GST," according to Eric Pop, an assistant professor of electrical and computer engineering at the University of Illinois.
The Exeter team used an optical device that employed memflectors, which are the optical analog of memristors.
Memristors are passive two-terminal electric components in which the resistance depends on the directional flow of the current. They "remember" the last resistance they had and, when a charge is put through them again, their resistance will be what it was when the charge last flowed through the circuit.
Memflectors are devices whose optical reflectivity is determined by the charge sent through them, making them the optical analogs of memristors.
Whereas memristors store information as resistance, memflectors store it as a change in optical reflectance, Pop told TechNewsWorld.
Volts Beat Out the Light
The demonstration that PCMs can be used to make artificial neurons and synapses means that an artificial system made entirely from phase-change devices could potentially learn and process information the way our brains do, the University of Exeter researchers stated.
This project focused on the performance of a single phase-change cell. The university's next step will be to build systems of interconnected cells that will be able to "learn" to perform simple tasks such as the identification of selected objects and patterns, it said.
However, we'll have to go beyond using memflectors if we want to get to brain-like computing successfully, the University of Illinois' Pop stated.
The next advance for a brain-on-a-chip will be "to demonstrate such information processing using electrical, rather than optical, pulses," Pop remarked.